CRET_Experiment.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N2KhdY8LcmIH"
      },
      "source": [
        "# Imports"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X0O6bVUUcwxR"
      },
      "source": [
        "## Required Imports"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 9,
      "metadata": {
        "id": "0h927EScc1tA"
      },
      "outputs": [],
      "source": [
        "import os\n",
        "import json\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "from transformers import BertModel, BertTokenizer\n",
        "import torchvision.models as models\n",
        "from typing import List, Tuple\n",
        "import cv2\n",
        "import numpy as np\n",
        "from tqdm import tqdm  # for progress bars"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z__uc013U3N5"
      },
      "source": [
        "## Video Encoder"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 10,
      "metadata": {
        "id": "zqNQQEH4U4jl"
      },
      "outputs": [],
      "source": [
        "class VideoEncoder(nn.Module):\n",
        "    def __init__(self, d_model=768):\n",
        "        super().__init__()\n",
        "        self.spatial_encoder = models.resnet50(pretrained=True)\n",
        "        self.spatial_encoder = nn.Sequential(*list(self.spatial_encoder.children())[:-1])\n",
        "        self.projection = nn.Linear(2048, d_model)\n",
        "\n",
        "        # Modified transformer encoder\n",
        "        encoder_layer = nn.TransformerEncoderLayer(\n",
        "            d_model=d_model,\n",
        "            nhead=8,\n",
        "            dim_feedforward=2048,\n",
        "            batch_first=True  # Important for dimension handling\n",
        "        )\n",
        "        self.temporal_encoder = nn.TransformerEncoder(encoder_layer, num_layers=3)\n",
        "\n",
        "    def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        B, F, C, H, W = x.shape\n",
        "\n",
        "        # Process each frame\n",
        "        x = x.view(B * F, C, H, W)\n",
        "        spatial_features = self.spatial_encoder(x)\n",
        "        spatial_features = spatial_features.squeeze(-1).squeeze(-1)  # Remove spatial dimensions\n",
        "        spatial_features = self.projection(spatial_features)\n",
        "\n",
        "        # Reshape for temporal processing\n",
        "        spatial_features = spatial_features.view(B, F, -1)  # [batch, frames, features]\n",
        "\n",
        "        # Temporal encoding (now with batch_first=True)\n",
        "        temporal_features = self.temporal_encoder(spatial_features)\n",
        "\n",
        "        return temporal_features, spatial_features"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Dylz_SSLVWYA"
      },
      "source": [
        "## CCM Module"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 11,
      "metadata": {
        "id": "nGuZ1OCnVYAV"
      },
      "outputs": [],
      "source": [
        "class CCMModule(nn.Module):\n",
        "    def __init__(self, d_model=768, num_heads=8, num_queries=8):\n",
        "        super().__init__()\n",
        "        self.num_queries = num_queries\n",
        "        self.query_centers = nn.Parameter(torch.randn(num_queries, d_model))\n",
        "\n",
        "        # Modified decoder layer\n",
        "        self.decoder = nn.TransformerDecoderLayer(\n",
        "            d_model=d_model,\n",
        "            nhead=num_heads,\n",
        "            dim_feedforward=2048,\n",
        "            batch_first=True  # Important for dimension handling\n",
        "        )\n",
        "\n",
        "    def forward(self, video_features: torch.Tensor, text_features: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        batch_size = video_features.size(0)\n",
        "\n",
        "        # Expand queries for batch dimension\n",
        "        query = self.query_centers.unsqueeze(0).expand(batch_size, -1, -1)\n",
        "\n",
        "        # Apply decoder to both modalities\n",
        "        video_aligned = self.decoder(query, video_features)\n",
        "        text_aligned = self.decoder(query, text_features)\n",
        "\n",
        "        return video_aligned, text_aligned"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eBjhimDdVZdv"
      },
      "source": [
        "## CRET Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 12,
      "metadata": {
        "id": "OjaCw6nNVcvF"
      },
      "outputs": [],
      "source": [
        "class CRET(nn.Module):\n",
        "    def __init__(self, d_model=768):\n",
        "        super().__init__()\n",
        "        self.video_encoder = VideoEncoder(d_model)\n",
        "        self.text_encoder = BertModel.from_pretrained('bert-base-uncased')\n",
        "        self.ccm = CCMModule(d_model)\n",
        "\n",
        "    def forward(self, video: torch.Tensor, text_ids: torch.Tensor, text_mask: torch.Tensor):\n",
        "        temporal_features, spatial_features = self.video_encoder(video)\n",
        "        text_outputs = self.text_encoder(text_ids, attention_mask=text_mask)\n",
        "        text_features = text_outputs.last_hidden_state\n",
        "        video_aligned, text_aligned = self.ccm(spatial_features, text_features)\n",
        "        video_global = temporal_features.mean(1)\n",
        "        text_global = text_outputs.pooler_output\n",
        "        return {\n",
        "            'video_aligned': video_aligned,\n",
        "            'text_aligned': text_aligned,\n",
        "            'video_global': video_global,\n",
        "            'text_global': text_global\n",
        "        }"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9ytAhfAoVgWv"
      },
      "source": [
        "## GEES Loss"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 13,
      "metadata": {
        "id": "xto08ysHVi_a"
      },
      "outputs": [],
      "source": [
        "class GEESLoss(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "\n",
        "    def forward(self, video_features: torch.Tensor, text_features: torch.Tensor):\n",
        "        # Make sure video_features has the right shape [batch_size, seq_len, features]\n",
        "        if len(video_features.shape) == 2:\n",
        "            video_features = video_features.unsqueeze(1)\n",
        "\n",
        "        # Compute mean over sequence dimension (if exists)\n",
        "        if len(video_features.shape) == 3:\n",
        "            video_mean = video_features.mean(1)\n",
        "        else:\n",
        "            video_mean = video_features\n",
        "\n",
        "        # Compute covariance\n",
        "        batch_size = video_features.size(0)\n",
        "        feature_dim = video_features.size(-1)\n",
        "\n",
        "        # Reshape video features for batch matrix multiplication\n",
        "        video_features_reshaped = video_features.view(batch_size, -1, feature_dim)\n",
        "\n",
        "        # Compute covariance for each item in the batch\n",
        "        video_cov = torch.bmm(video_features_reshaped.transpose(1, 2),\n",
        "                            video_features_reshaped) / video_features_reshaped.size(1)\n",
        "\n",
        "        # Compute similarity scores\n",
        "        sim = torch.mm(text_features, video_mean.t())\n",
        "\n",
        "        # Compute covariance term\n",
        "        cov_term = 0.5 * torch.bmm(\n",
        "            torch.bmm(text_features.unsqueeze(1), video_cov),\n",
        "            text_features.unsqueeze(-1)\n",
        "        ).squeeze()\n",
        "\n",
        "        # Compute logits and loss\n",
        "        logits = sim + cov_term\n",
        "        labels = torch.arange(batch_size).to(video_features.device)\n",
        "\n",
        "        return F.cross_entropy(logits, labels)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xgW8wCKJWHLT"
      },
      "source": [
        "## Dataset"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 14,
      "metadata": {
        "id": "z_RNeZWIWJ-8"
      },
      "outputs": [],
      "source": [
        "class MSRVTTDataset(Dataset):\n",
        "    def __init__(self, video_paths, captions, tokenizer):\n",
        "        self.video_paths = video_paths\n",
        "        self.captions = captions\n",
        "        self.tokenizer = tokenizer\n",
        "\n",
        "        # Add cv2 import inside the class\n",
        "        import cv2\n",
        "        self.cv2 = cv2\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.video_paths)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        video = self.load_video(self.video_paths[idx])\n",
        "        caption = self.captions[idx]\n",
        "        tokens = self.tokenizer(\n",
        "            caption,\n",
        "            padding='max_length',\n",
        "            truncation=True,\n",
        "            max_length=128,\n",
        "            return_tensors='pt'\n",
        "        )\n",
        "        return video, tokens.input_ids.squeeze(0), tokens.attention_mask.squeeze(0)\n",
        "\n",
        "    def load_video(self, path):\n",
        "        import numpy as np\n",
        "        frames = []\n",
        "\n",
        "        # Load video with OpenCV\n",
        "        cap = self.cv2.VideoCapture(path)\n",
        "\n",
        "        # Sample 4 frames as mentioned in the paper\n",
        "        total_frames = int(cap.get(self.cv2.CAP_PROP_FRAME_COUNT))\n",
        "        indices = np.linspace(0, total_frames-1, 4, dtype=int)\n",
        "\n",
        "        for frame_idx in indices:\n",
        "            cap.set(self.cv2.CAP_PROP_POS_FRAMES, frame_idx)\n",
        "            ret, frame = cap.read()\n",
        "            if ret:\n",
        "                # Convert BGR to RGB\n",
        "                frame = self.cv2.cvtColor(frame, self.cv2.COLOR_BGR2RGB)\n",
        "                # Resize to 224x224\n",
        "                frame = self.cv2.resize(frame, (224, 224))\n",
        "                # Convert to tensor and normalize\n",
        "                frame = torch.from_numpy(frame).permute(2, 0, 1).float() / 255.0\n",
        "                frames.append(frame)\n",
        "\n",
        "        cap.release()\n",
        "\n",
        "        # Stack frames\n",
        "        if len(frames) == 0:  # Handle empty video case\n",
        "            return torch.zeros((4, 3, 224, 224))\n",
        "\n",
        "        video_tensor = torch.stack(frames)\n",
        "        return video_tensor"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "D10oN3Tsxj0p"
      },
      "source": [
        "## Training Function"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 15,
      "metadata": {
        "id": "DN9VaM55xnCi"
      },
      "outputs": [],
      "source": [
        "def train_cret(model, train_loader, num_epochs=2, device='cpu'):\n",
        "    print(f\"Training on device: {device}\")\n",
        "    model = model.to(device)\n",
        "    gees_loss = GEESLoss()\n",
        "    optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        model.train()\n",
        "        total_loss = 0\n",
        "\n",
        "        for batch_idx, (videos, text_ids, text_mask) in enumerate(train_loader):\n",
        "            try:\n",
        "                # Move to device\n",
        "                videos = videos.to(device).float()\n",
        "                text_ids = text_ids.to(device)\n",
        "                text_mask = text_mask.to(device)\n",
        "\n",
        "                # Forward pass\n",
        "                outputs = model(videos, text_ids, text_mask)\n",
        "\n",
        "                # Print shapes for debugging\n",
        "                if batch_idx == 0:\n",
        "                    print(f\"Video global shape: {outputs['video_global'].shape}\")\n",
        "                    print(f\"Text global shape: {outputs['text_global'].shape}\")\n",
        "                    print(f\"Video aligned shape: {outputs['video_aligned'].shape}\")\n",
        "                    print(f\"Text aligned shape: {outputs['text_aligned'].shape}\")\n",
        "\n",
        "                # Compute losses\n",
        "                ccm_loss = F.cosine_embedding_loss(\n",
        "                    outputs['video_aligned'].view(-1, 768),\n",
        "                    outputs['text_aligned'].view(-1, 768),\n",
        "                    torch.ones(outputs['video_aligned'].size(0) * outputs['video_aligned'].size(1)).to(device)\n",
        "                )\n",
        "\n",
        "                gees_loss_val = gees_loss(outputs['video_global'], outputs['text_global'])\n",
        "                loss = ccm_loss + gees_loss_val\n",
        "\n",
        "                # Backward pass\n",
        "                optimizer.zero_grad()\n",
        "                loss.backward()\n",
        "                optimizer.step()\n",
        "\n",
        "                total_loss += loss.item()\n",
        "\n",
        "                if batch_idx % 10 == 0:\n",
        "                    print(f'Epoch: {epoch}, Batch: {batch_idx}, Loss: {loss.item():.4f}')\n",
        "\n",
        "            except Exception as e:\n",
        "                print(f\"Error in batch {batch_idx}: {str(e)}\")\n",
        "                print(f\"Shapes:\")\n",
        "                for k, v in outputs.items():\n",
        "                    print(f\"{k}: {v.shape}\")\n",
        "                continue\n",
        "\n",
        "        avg_loss = total_loss / len(train_loader)\n",
        "        print(f'Epoch {epoch} completed, Average Loss: {avg_loss:.4f}')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tDfiItClx_r5"
      },
      "source": [
        "## Main Execution"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 16,
      "metadata": {
        "id": "J0OH_udvyEIm"
      },
      "outputs": [],
      "source": [
        "def main():\n",
        "    # Initialize tokenizer\n",
        "    tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
        "\n",
        "    # Load MSRVTT dataset\n",
        "    msrvtt_root = './data/MSRVTT'\n",
        "    # Create the videos directory if it doesn't exist\n",
        "    os.makedirs(os.path.join(msrvtt_root, 'videos'), exist_ok=True)\n",
        "\n",
        "    # Download video data and captions if they don't exist\n",
        "    if not os.listdir(os.path.join(msrvtt_root, 'videos')):\n",
        "        print(\"Downloading video data...\")\n",
        "        # Replace <updated_video_data_download_link> with the actual download link\n",
        "        os.system('wget -P ./data/MSRVTT/videos <updated_video_data_download_link>')\n",
        "    if not os.path.exists(os.path.join(msrvtt_root, 'captions.json')):\n",
        "        print(\"Downloading captions...\")\n",
        "        os.system('wget -P ./data/MSRVTT https://www.robots.ox.ac.uk/~vgg/research/collaborative-experts/data/msrvtt_data.json')\n",
        "\n",
        "    video_paths = [os.path.join(msrvtt_root, 'videos', file) for file in os.listdir(os.path.join(msrvtt_root, 'videos')) if file.endswith('.mp4')]\n",
        "\n",
        "    with open(os.path.join(msrvtt_root, 'captions.json'), 'r') as f:\n",
        "        captions_data = json.load(f)\n",
        "\n",
        "    captions = [captions_data[video_id[:-4]] for video_id in os.listdir(os.path.join(msrvtt_root, 'videos')) if video_id.endswith('.mp4')]\n",
        "    captions = [[caption['caption'] for caption in captions_per_video] for captions_per_video in captions]\n",
        "    # Create dataset and dataloader\n",
        "    dataset = MSRVTTDataset(video_paths, captions, tokenizer)\n",
        "    train_loader = DataLoader(dataset, batch_size=32, shuffle=True)\n",
        "\n",
        "    # Initialize model\n",
        "    model = CRET()\n",
        "\n",
        "    # Train model\n",
        "    train_cret(model, train_loader)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 17,
      "metadata": {
        "id": "2j4-BKzM9ADH"
      },
      "outputs": [],
      "source": [
        "def main():\n",
        "    # Initialize tokenizer\n",
        "    tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
        "\n",
        "    # Set up paths for your Colab environment\n",
        "    msrvtt_root = '/content/MSRVTT'  # Adjust this to your actual path\n",
        "\n",
        "    # Get video paths and captions\n",
        "    video_paths = [os.path.join(msrvtt_root, 'videos', file)\n",
        "                  for file in os.listdir(os.path.join(msrvtt_root, 'videos'))\n",
        "                  if file.endswith('.mp4')]\n",
        "\n",
        "    # Load captions\n",
        "    with open(os.path.join(msrvtt_root, 'annotation', 'captions.json'), 'r') as f:\n",
        "        captions_data = json.load(f)\n",
        "\n",
        "    # Create dataset and dataloader\n",
        "    dataset = MSRVTTDataset(video_paths[:100], captions[:100], tokenizer)  # Start with 100 videos for testing\n",
        "    train_loader = DataLoader(dataset, batch_size=8, shuffle=True)  # Smaller batch size for Colab\n",
        "\n",
        "    # Initialize model\n",
        "    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "    model = CRET().to(device)\n",
        "\n",
        "    # Train model\n",
        "    train_cret(model, train_loader, num_epochs=5, device=device)  # Start with 5 epochs for testing"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 18,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "UxRjTtBt9A5R",
        "outputId": "9540d00f-1a77-4cd9-c794-41df191a347b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "GPU available: False\n"
          ]
        }
      ],
      "source": [
        "import torch\n",
        "print(f\"GPU available: {torch.cuda.is_available()}\")\n",
        "if torch.cuda.is_available():\n",
        "    print(f\"GPU device: {torch.cuda.get_device_name(0)}\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 19,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "K74G13_BFB8S",
        "outputId": "55b5f2c1-cc99-4c93-aed2-78f422c57ac2"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Current directory: /content\n",
            "\n",
            "Contents of current directory:\n",
            "MSRVTT\tsample_data\n",
            "\n",
            "Contents of /content directory:\n",
            "MSRVTT\tsample_data\n"
          ]
        }
      ],
      "source": [
        "# Check current directory and list files\n",
        "print(\"Current directory:\", os.getcwd())\n",
        "print(\"\\nContents of current directory:\")\n",
        "!ls\n",
        "\n",
        "print(\"\\nContents of /content directory:\")\n",
        "!ls /content"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 20,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZzTwqzHzFI1E",
        "outputId": "507bda2b-f208-49b8-eba9-02e9cd6e2ab1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Contents of MSRVTT folder:\n"
          ]
        }
      ],
      "source": [
        "print(\"Contents of MSRVTT folder:\")\n",
        "!ls /content/MSRVTT"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 33,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "dp-Xcqv69C33",
        "outputId": "84a5f2d0-c8c4-4588-ae9a-8162dc0dfc0f"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Testing dataset loading...\n",
            "\n",
            "Checking contents of MSRVTT directory:\n",
            "test_annotations.json  video249.mp4  video39.mp4   video54.mp4\t video6.mp4    video850.mp4\n",
            "video0.mp4\t       video24.mp4   video3.mp4    video550.mp4  video700.mp4  video851.mp4\n",
            "video1000.mp4\t       video250.mp4  video400.mp4  video551.mp4  video701.mp4  video852.mp4\n",
            "video100.mp4\t       video251.mp4  video401.mp4  video552.mp4  video702.mp4  video853.mp4\n",
            "video101.mp4\t       video252.mp4  video402.mp4  video553.mp4  video703.mp4  video854.mp4\n",
            "video102.mp4\t       video253.mp4  video403.mp4  video554.mp4  video704.mp4  video855.mp4\n",
            "video103.mp4\t       video254.mp4  video404.mp4  video555.mp4  video705.mp4  video856.mp4\n",
            "video104.mp4\t       video255.mp4  video405.mp4  video556.mp4  video706.mp4  video857.mp4\n",
            "video105.mp4\t       video256.mp4  video406.mp4  video557.mp4  video707.mp4  video858.mp4\n",
            "video106.mp4\t       video257.mp4  video407.mp4  video558.mp4  video708.mp4  video859.mp4\n",
            "video107.mp4\t       video258.mp4  video408.mp4  video559.mp4  video709.mp4  video85.mp4\n",
            "video108.mp4\t       video259.mp4  video409.mp4  video55.mp4\t video70.mp4   video860.mp4\n",
            "video109.mp4\t       video25.mp4   video40.mp4   video560.mp4  video710.mp4  video861.mp4\n",
            "video10.mp4\t       video260.mp4  video410.mp4  video561.mp4  video711.mp4  video862.mp4\n",
            "video110.mp4\t       video261.mp4  video411.mp4  video562.mp4  video712.mp4  video863.mp4\n",
            "video111.mp4\t       video262.mp4  video412.mp4  video563.mp4  video713.mp4  video864.mp4\n",
            "video112.mp4\t       video263.mp4  video413.mp4  video564.mp4  video714.mp4  video865.mp4\n",
            "video113.mp4\t       video264.mp4  video414.mp4  video565.mp4  video715.mp4  video866.mp4\n",
            "video114.mp4\t       video265.mp4  video415.mp4  video566.mp4  video716.mp4  video867.mp4\n",
            "video115.mp4\t       video266.mp4  video416.mp4  video567.mp4  video717.mp4  video868.mp4\n",
            "video116.mp4\t       video267.mp4  video417.mp4  video568.mp4  video718.mp4  video869.mp4\n",
            "video117.mp4\t       video268.mp4  video418.mp4  video569.mp4  video719.mp4  video86.mp4\n",
            "video118.mp4\t       video269.mp4  video419.mp4  video56.mp4\t video71.mp4   video870.mp4\n",
            "video119.mp4\t       video26.mp4   video41.mp4   video570.mp4  video720.mp4  video871.mp4\n",
            "video11.mp4\t       video270.mp4  video420.mp4  video571.mp4  video721.mp4  video872.mp4\n",
            "video120.mp4\t       video271.mp4  video421.mp4  video572.mp4  video722.mp4  video873.mp4\n",
            "video121.mp4\t       video272.mp4  video422.mp4  video573.mp4  video723.mp4  video874.mp4\n",
            "video122.mp4\t       video273.mp4  video423.mp4  video574.mp4  video724.mp4  video875.mp4\n",
            "video123.mp4\t       video274.mp4  video424.mp4  video575.mp4  video725.mp4  video876.mp4\n",
            "video124.mp4\t       video275.mp4  video425.mp4  video576.mp4  video726.mp4  video877.mp4\n",
            "video125.mp4\t       video276.mp4  video426.mp4  video577.mp4  video727.mp4  video878.mp4\n",
            "video126.mp4\t       video277.mp4  video427.mp4  video578.mp4  video728.mp4  video879.mp4\n",
            "video127.mp4\t       video278.mp4  video428.mp4  video579.mp4  video729.mp4  video87.mp4\n",
            "video128.mp4\t       video279.mp4  video429.mp4  video57.mp4\t video72.mp4   video880.mp4\n",
            "video129.mp4\t       video27.mp4   video42.mp4   video580.mp4  video730.mp4  video881.mp4\n",
            "video12.mp4\t       video280.mp4  video430.mp4  video581.mp4  video731.mp4  video882.mp4\n",
            "video130.mp4\t       video281.mp4  video431.mp4  video582.mp4  video732.mp4  video883.mp4\n",
            "video131.mp4\t       video282.mp4  video432.mp4  video583.mp4  video733.mp4  video884.mp4\n",
            "video132.mp4\t       video283.mp4  video433.mp4  video584.mp4  video734.mp4  video885.mp4\n",
            "video133.mp4\t       video284.mp4  video434.mp4  video585.mp4  video735.mp4  video886.mp4\n",
            "video134.mp4\t       video285.mp4  video435.mp4  video586.mp4  video736.mp4  video887.mp4\n",
            "video135.mp4\t       video286.mp4  video436.mp4  video587.mp4  video737.mp4  video888.mp4\n",
            "video136.mp4\t       video287.mp4  video437.mp4  video588.mp4  video738.mp4  video889.mp4\n",
            "video137.mp4\t       video288.mp4  video438.mp4  video589.mp4  video739.mp4  video88.mp4\n",
            "video138.mp4\t       video289.mp4  video439.mp4  video58.mp4\t video73.mp4   video890.mp4\n",
            "video139.mp4\t       video28.mp4   video43.mp4   video590.mp4  video740.mp4  video891.mp4\n",
            "video13.mp4\t       video290.mp4  video440.mp4  video591.mp4  video741.mp4  video892.mp4\n",
            "video140.mp4\t       video291.mp4  video441.mp4  video592.mp4  video742.mp4  video893.mp4\n",
            "video141.mp4\t       video292.mp4  video442.mp4  video593.mp4  video743.mp4  video894.mp4\n",
            "video142.mp4\t       video293.mp4  video443.mp4  video594.mp4  video744.mp4  video895.mp4\n",
            "video143.mp4\t       video294.mp4  video444.mp4  video595.mp4  video745.mp4  video896.mp4\n",
            "video144.mp4\t       video295.mp4  video445.mp4  video596.mp4  video746.mp4  video897.mp4\n",
            "video145.mp4\t       video296.mp4  video446.mp4  video597.mp4  video747.mp4  video898.mp4\n",
            "video146.mp4\t       video297.mp4  video447.mp4  video598.mp4  video748.mp4  video899.mp4\n",
            "video147.mp4\t       video298.mp4  video448.mp4  video599.mp4  video749.mp4  video89.mp4\n",
            "video148.mp4\t       video299.mp4  video449.mp4  video59.mp4\t video74.mp4   video8.mp4\n",
            "video149.mp4\t       video29.mp4   video44.mp4   video5.mp4\t video750.mp4  video900.mp4\n",
            "video14.mp4\t       video2.mp4    video450.mp4  video600.mp4  video751.mp4  video901.mp4\n",
            "video150.mp4\t       video300.mp4  video451.mp4  video601.mp4  video752.mp4  video902.mp4\n",
            "video151.mp4\t       video301.mp4  video452.mp4  video602.mp4  video753.mp4  video903.mp4\n",
            "video152.mp4\t       video302.mp4  video453.mp4  video603.mp4  video754.mp4  video904.mp4\n",
            "video153.mp4\t       video303.mp4  video454.mp4  video604.mp4  video755.mp4  video905.mp4\n",
            "video154.mp4\t       video304.mp4  video455.mp4  video605.mp4  video756.mp4  video906.mp4\n",
            "video155.mp4\t       video305.mp4  video456.mp4  video606.mp4  video757.mp4  video907.mp4\n",
            "video156.mp4\t       video306.mp4  video457.mp4  video607.mp4  video758.mp4  video908.mp4\n",
            "video157.mp4\t       video307.mp4  video458.mp4  video608.mp4  video759.mp4  video909.mp4\n",
            "video158.mp4\t       video308.mp4  video459.mp4  video609.mp4  video75.mp4   video90.mp4\n",
            "video159.mp4\t       video309.mp4  video45.mp4   video60.mp4\t video760.mp4  video910.mp4\n",
            "video15.mp4\t       video30.mp4   video460.mp4  video610.mp4  video761.mp4  video911.mp4\n",
            "video160.mp4\t       video310.mp4  video461.mp4  video611.mp4  video762.mp4  video912.mp4\n",
            "video161.mp4\t       video311.mp4  video462.mp4  video612.mp4  video763.mp4  video913.mp4\n",
            "video162.mp4\t       video312.mp4  video463.mp4  video613.mp4  video764.mp4  video914.mp4\n",
            "video163.mp4\t       video313.mp4  video464.mp4  video614.mp4  video765.mp4  video915.mp4\n",
            "video164.mp4\t       video314.mp4  video465.mp4  video615.mp4  video766.mp4  video916.mp4\n",
            "video165.mp4\t       video315.mp4  video466.mp4  video616.mp4  video767.mp4  video917.mp4\n",
            "video166.mp4\t       video316.mp4  video467.mp4  video617.mp4  video768.mp4  video918.mp4\n",
            "video167.mp4\t       video317.mp4  video468.mp4  video618.mp4  video769.mp4  video919.mp4\n",
            "video168.mp4\t       video318.mp4  video469.mp4  video619.mp4  video76.mp4   video91.mp4\n",
            "video169.mp4\t       video319.mp4  video46.mp4   video61.mp4\t video770.mp4  video920.mp4\n",
            "video16.mp4\t       video31.mp4   video470.mp4  video620.mp4  video771.mp4  video921.mp4\n",
            "video170.mp4\t       video320.mp4  video471.mp4  video621.mp4  video772.mp4  video922.mp4\n",
            "video171.mp4\t       video321.mp4  video472.mp4  video622.mp4  video773.mp4  video923.mp4\n",
            "video172.mp4\t       video322.mp4  video473.mp4  video623.mp4  video774.mp4  video924.mp4\n",
            "video173.mp4\t       video323.mp4  video474.mp4  video624.mp4  video775.mp4  video925.mp4\n",
            "video174.mp4\t       video324.mp4  video475.mp4  video625.mp4  video776.mp4  video926.mp4\n",
            "video175.mp4\t       video325.mp4  video476.mp4  video626.mp4  video777.mp4  video927.mp4\n",
            "video176.mp4\t       video326.mp4  video477.mp4  video627.mp4  video778.mp4  video928.mp4\n",
            "video177.mp4\t       video327.mp4  video478.mp4  video628.mp4  video779.mp4  video929.mp4\n",
            "video178.mp4\t       video328.mp4  video479.mp4  video629.mp4  video77.mp4   video92.mp4\n",
            "video179.mp4\t       video329.mp4  video47.mp4   video62.mp4\t video780.mp4  video930.mp4\n",
            "video17.mp4\t       video32.mp4   video480.mp4  video630.mp4  video781.mp4  video931.mp4\n",
            "video180.mp4\t       video330.mp4  video481.mp4  video631.mp4  video782.mp4  video932.mp4\n",
            "video181.mp4\t       video331.mp4  video482.mp4  video632.mp4  video783.mp4  video933.mp4\n",
            "video182.mp4\t       video332.mp4  video483.mp4  video633.mp4  video784.mp4  video934.mp4\n",
            "video183.mp4\t       video333.mp4  video484.mp4  video634.mp4  video785.mp4  video935.mp4\n",
            "video184.mp4\t       video334.mp4  video485.mp4  video635.mp4  video786.mp4  video936.mp4\n",
            "video185.mp4\t       video335.mp4  video486.mp4  video636.mp4  video787.mp4  video937.mp4\n",
            "video186.mp4\t       video336.mp4  video487.mp4  video637.mp4  video788.mp4  video938.mp4\n",
            "video187.mp4\t       video337.mp4  video488.mp4  video638.mp4  video789.mp4  video939.mp4\n",
            "video188.mp4\t       video338.mp4  video489.mp4  video639.mp4  video78.mp4   video93.mp4\n",
            "video189.mp4\t       video339.mp4  video48.mp4   video63.mp4\t video790.mp4  video940.mp4\n",
            "video18.mp4\t       video33.mp4   video490.mp4  video640.mp4  video791.mp4  video941.mp4\n",
            "video190.mp4\t       video340.mp4  video491.mp4  video641.mp4  video792.mp4  video942.mp4\n",
            "video191.mp4\t       video341.mp4  video492.mp4  video642.mp4  video793.mp4  video943.mp4\n",
            "video192.mp4\t       video342.mp4  video493.mp4  video643.mp4  video794.mp4  video944.mp4\n",
            "video193.mp4\t       video343.mp4  video494.mp4  video644.mp4  video795.mp4  video945.mp4\n",
            "video194.mp4\t       video344.mp4  video495.mp4  video645.mp4  video796.mp4  video946.mp4\n",
            "video195.mp4\t       video345.mp4  video496.mp4  video646.mp4  video797.mp4  video947.mp4\n",
            "video196.mp4\t       video346.mp4  video497.mp4  video647.mp4  video798.mp4  video948.mp4\n",
            "video197.mp4\t       video347.mp4  video498.mp4  video648.mp4  video799.mp4  video949.mp4\n",
            "video198.mp4\t       video348.mp4  video499.mp4  video649.mp4  video79.mp4   video94.mp4\n",
            "video199.mp4\t       video349.mp4  video49.mp4   video64.mp4\t video7.mp4    video950.mp4\n",
            "video19.mp4\t       video34.mp4   video4.mp4    video650.mp4  video800.mp4  video951.mp4\n",
            "video1.mp4\t       video350.mp4  video500.mp4  video651.mp4  video801.mp4  video952.mp4\n",
            "video200.mp4\t       video351.mp4  video501.mp4  video652.mp4  video802.mp4  video953.mp4\n",
            "video201.mp4\t       video352.mp4  video502.mp4  video653.mp4  video803.mp4  video954.mp4\n",
            "video202.mp4\t       video353.mp4  video503.mp4  video654.mp4  video804.mp4  video955.mp4\n",
            "video203.mp4\t       video354.mp4  video504.mp4  video655.mp4  video805.mp4  video956.mp4\n",
            "video204.mp4\t       video355.mp4  video505.mp4  video656.mp4  video806.mp4  video957.mp4\n",
            "video205.mp4\t       video356.mp4  video506.mp4  video657.mp4  video807.mp4  video958.mp4\n",
            "video206.mp4\t       video357.mp4  video507.mp4  video658.mp4  video808.mp4  video959.mp4\n",
            "video207.mp4\t       video358.mp4  video508.mp4  video659.mp4  video809.mp4  video95.mp4\n",
            "video208.mp4\t       video359.mp4  video509.mp4  video65.mp4\t video80.mp4   video960.mp4\n",
            "video209.mp4\t       video35.mp4   video50.mp4   video660.mp4  video810.mp4  video961.mp4\n",
            "video20.mp4\t       video360.mp4  video510.mp4  video661.mp4  video811.mp4  video962.mp4\n",
            "video210.mp4\t       video361.mp4  video511.mp4  video662.mp4  video812.mp4  video963.mp4\n",
            "video211.mp4\t       video362.mp4  video512.mp4  video663.mp4  video813.mp4  video964.mp4\n",
            "video212.mp4\t       video363.mp4  video513.mp4  video664.mp4  video814.mp4  video965.mp4\n",
            "video213.mp4\t       video364.mp4  video514.mp4  video665.mp4  video815.mp4  video966.mp4\n",
            "video214.mp4\t       video365.mp4  video515.mp4  video666.mp4  video816.mp4  video967.mp4\n",
            "video215.mp4\t       video366.mp4  video516.mp4  video667.mp4  video817.mp4  video968.mp4\n",
            "video216.mp4\t       video367.mp4  video517.mp4  video668.mp4  video818.mp4  video969.mp4\n",
            "video217.mp4\t       video368.mp4  video518.mp4  video669.mp4  video819.mp4  video96.mp4\n",
            "video218.mp4\t       video369.mp4  video519.mp4  video66.mp4\t video81.mp4   video970.mp4\n",
            "video219.mp4\t       video36.mp4   video51.mp4   video670.mp4  video820.mp4  video971.mp4\n",
            "video21.mp4\t       video370.mp4  video520.mp4  video671.mp4  video821.mp4  video972.mp4\n",
            "video220.mp4\t       video371.mp4  video521.mp4  video672.mp4  video822.mp4  video973.mp4\n",
            "video221.mp4\t       video372.mp4  video522.mp4  video673.mp4  video823.mp4  video974.mp4\n",
            "video222.mp4\t       video373.mp4  video523.mp4  video674.mp4  video824.mp4  video975.mp4\n",
            "video223.mp4\t       video374.mp4  video524.mp4  video675.mp4  video825.mp4  video976.mp4\n",
            "video224.mp4\t       video375.mp4  video525.mp4  video676.mp4  video826.mp4  video977.mp4\n",
            "video225.mp4\t       video376.mp4  video526.mp4  video677.mp4  video827.mp4  video978.mp4\n",
            "video226.mp4\t       video377.mp4  video527.mp4  video678.mp4  video828.mp4  video979.mp4\n",
            "video227.mp4\t       video378.mp4  video528.mp4  video679.mp4  video829.mp4  video97.mp4\n",
            "video228.mp4\t       video379.mp4  video529.mp4  video67.mp4\t video82.mp4   video980.mp4\n",
            "video229.mp4\t       video37.mp4   video52.mp4   video680.mp4  video830.mp4  video981.mp4\n",
            "video22.mp4\t       video380.mp4  video530.mp4  video681.mp4  video831.mp4  video982.mp4\n",
            "video230.mp4\t       video381.mp4  video531.mp4  video682.mp4  video832.mp4  video983.mp4\n",
            "video231.mp4\t       video382.mp4  video532.mp4  video683.mp4  video833.mp4  video984.mp4\n",
            "video232.mp4\t       video383.mp4  video533.mp4  video684.mp4  video834.mp4  video985.mp4\n",
            "video233.mp4\t       video384.mp4  video534.mp4  video685.mp4  video835.mp4  video986.mp4\n",
            "video234.mp4\t       video385.mp4  video535.mp4  video686.mp4  video836.mp4  video987.mp4\n",
            "video235.mp4\t       video386.mp4  video536.mp4  video687.mp4  video837.mp4  video988.mp4\n",
            "video236.mp4\t       video387.mp4  video537.mp4  video688.mp4  video838.mp4  video989.mp4\n",
            "video237.mp4\t       video388.mp4  video538.mp4  video689.mp4  video839.mp4  video98.mp4\n",
            "video238.mp4\t       video389.mp4  video539.mp4  video68.mp4\t video83.mp4   video990.mp4\n",
            "video239.mp4\t       video38.mp4   video53.mp4   video690.mp4  video840.mp4  video991.mp4\n",
            "video23.mp4\t       video390.mp4  video540.mp4  video691.mp4  video841.mp4  video992.mp4\n",
            "video240.mp4\t       video391.mp4  video541.mp4  video692.mp4  video842.mp4  video993.mp4\n",
            "video241.mp4\t       video392.mp4  video542.mp4  video693.mp4  video843.mp4  video996.mp4\n",
            "video242.mp4\t       video393.mp4  video543.mp4  video694.mp4  video844.mp4  video997.mp4\n",
            "video243.mp4\t       video394.mp4  video544.mp4  video695.mp4  video845.mp4  video998.mp4\n",
            "video244.mp4\t       video395.mp4  video545.mp4  video696.mp4  video846.mp4  video999.mp4\n",
            "video245.mp4\t       video396.mp4  video546.mp4  video697.mp4  video847.mp4  video99.mp4\n",
            "video246.mp4\t       video397.mp4  video547.mp4  video698.mp4  video848.mp4  video9.mp4\n",
            "video247.mp4\t       video398.mp4  video548.mp4  video699.mp4  video849.mp4\n",
            "video248.mp4\t       video399.mp4  video549.mp4  video69.mp4\t video84.mp4\n",
            "\n",
            "Found 999 video files\n",
            "\n",
            "Trying to load a single video...\n",
            "\n",
            "Successfully loaded:\n",
            "Video tensor shape: torch.Size([4, 3, 224, 224])\n",
            "Text IDs shape: torch.Size([128])\n",
            "Text mask shape: torch.Size([128])\n"
          ]
        }
      ],
      "source": [
        "def test_dataset():\n",
        "    print(\"Testing dataset loading...\")\n",
        "\n",
        "    # Initialize tokenizer\n",
        "    tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
        "\n",
        "    # Set paths\n",
        "    msrvtt_root = '/content/MSRVTT'\n",
        "    print(f\"\\nChecking contents of MSRVTT directory:\")\n",
        "    !ls {msrvtt_root}\n",
        "\n",
        "    # Get video paths\n",
        "    video_paths = [os.path.join(msrvtt_root, f) for f in os.listdir(msrvtt_root) if f.endswith('.mp4')]\n",
        "    print(f\"\\nFound {len(video_paths)} video files\")\n",
        "\n",
        "    # For testing, let's use just one video with a dummy caption\n",
        "    test_dataset = MSRVTTDataset([video_paths[0]], ['a person is doing something'], tokenizer)\n",
        "    print(\"\\nTrying to load a single video...\")\n",
        "\n",
        "    try:\n",
        "        video, text_ids, text_mask = test_dataset[0]\n",
        "        print(f\"\\nSuccessfully loaded:\")\n",
        "        print(f\"Video tensor shape: {video.shape}\")\n",
        "        print(f\"Text IDs shape: {text_ids.shape}\")\n",
        "        print(f\"Text mask shape: {text_mask.shape}\")\n",
        "    except Exception as e:\n",
        "        print(f\"Error loading video: {e}\")\n",
        "\n",
        "test_dataset()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 34,
      "metadata": {
        "id": "g83S-QdaF9DZ"
      },
      "outputs": [],
      "source": [
        "def create_test_annotations():\n",
        "    video_files = [f for f in os.listdir('/content/MSRVTT') if f.endswith('.mp4')]\n",
        "    annotations = {}\n",
        "\n",
        "    for video_file in video_files:\n",
        "        video_id = video_file[:-4]  # Remove .mp4\n",
        "        annotations[video_id] = [{\n",
        "            'caption': f'This is a test caption for video {video_id}',\n",
        "            'video_id': video_id\n",
        "        }]\n",
        "\n",
        "    with open('/content/MSRVTT/test_annotations.json', 'w') as f:\n",
        "        json.dump(annotations, f)\n",
        "\n",
        "    print(\"Created test annotations file\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 35,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fQmbrvcDGHDA",
        "outputId": "d12c94a1-dc94-4cfe-f917-f21445a88114"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Created test annotations file\n"
          ]
        }
      ],
      "source": [
        "# Create test annotations\n",
        "def create_test_annotations():\n",
        "    video_files = [f for f in os.listdir('/content/MSRVTT') if f.endswith('.mp4')]\n",
        "    annotations = {}\n",
        "\n",
        "    for video_file in video_files:\n",
        "        video_id = video_file[:-4]  # Remove .mp4\n",
        "        annotations[video_id] = [{\n",
        "            'caption': f'This is a test caption for video {video_id}',\n",
        "            'video_id': video_id\n",
        "        }]\n",
        "\n",
        "    with open('/content/MSRVTT/test_annotations.json', 'w') as f:\n",
        "        json.dump(annotations, f)\n",
        "\n",
        "    print(\"Created test annotations file\")\n",
        "\n",
        "create_test_annotations()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 36,
      "metadata": {
        "id": "aRAUaiJEGIlT"
      },
      "outputs": [],
      "source": [
        "class MSRVTTDataset(Dataset):\n",
        "    def __init__(self, video_paths, captions, tokenizer):\n",
        "        self.video_paths = video_paths\n",
        "        self.captions = captions\n",
        "        self.tokenizer = tokenizer\n",
        "\n",
        "        # Add cv2 import inside the class\n",
        "        import cv2\n",
        "        self.cv2 = cv2\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.video_paths)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        video = self.load_video(self.video_paths[idx])\n",
        "        caption = self.captions[idx]\n",
        "        tokens = self.tokenizer(\n",
        "            caption,\n",
        "            padding='max_length',\n",
        "            truncation=True,\n",
        "            max_length=128,\n",
        "            return_tensors='pt'\n",
        "        )\n",
        "        return video, tokens.input_ids.squeeze(0), tokens.attention_mask.squeeze(0)\n",
        "\n",
        "    def load_video(self, path):\n",
        "        import numpy as np\n",
        "        frames = []\n",
        "\n",
        "        try:\n",
        "            # Load video with OpenCV\n",
        "            cap = self.cv2.VideoCapture(path)\n",
        "\n",
        "            # Get total frames\n",
        "            total_frames = int(cap.get(self.cv2.CAP_PROP_FRAME_COUNT))\n",
        "            if total_frames == 0:\n",
        "                # Return dummy frames if video is empty\n",
        "                return torch.zeros((4, 3, 224, 224))\n",
        "\n",
        "            # Sample 4 evenly spaced frames\n",
        "            indices = np.linspace(0, total_frames-1, 4, dtype=int)\n",
        "\n",
        "            for frame_idx in indices:\n",
        "                cap.set(self.cv2.CAP_PROP_POS_FRAMES, frame_idx)\n",
        "                ret, frame = cap.read()\n",
        "                if ret:\n",
        "                    # Convert BGR to RGB\n",
        "                    frame = self.cv2.cvtColor(frame, self.cv2.COLOR_BGR2RGB)\n",
        "                    # Resize to 224x224\n",
        "                    frame = self.cv2.resize(frame, (224, 224))\n",
        "                    # Convert to tensor and normalize\n",
        "                    frame = torch.from_numpy(frame).permute(2, 0, 1).float() / 255.0\n",
        "                    frames.append(frame)\n",
        "                else:\n",
        "                    # If frame read fails, append zero frame\n",
        "                    frames.append(torch.zeros((3, 224, 224)))\n",
        "\n",
        "            cap.release()\n",
        "\n",
        "            # Ensure we have exactly 4 frames\n",
        "            while len(frames) < 4:\n",
        "                frames.append(torch.zeros((3, 224, 224)))\n",
        "            frames = frames[:4]  # Take only first 4 if we somehow got more\n",
        "\n",
        "            # Stack frames\n",
        "            video_tensor = torch.stack(frames)\n",
        "            return video_tensor\n",
        "\n",
        "        except Exception as e:\n",
        "            print(f\"Error loading video {path}: {str(e)}\")\n",
        "            # Return dummy frames if loading fails\n",
        "            return torch.zeros((4, 3, 224, 224))"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "31kii1tqJi_y"
      },
      "source": [
        "## Monitor Training"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 37,
      "metadata": {
        "id": "SA_SCFQjJkiD"
      },
      "outputs": [],
      "source": [
        "from tqdm import tqdm  # For progress bars\n",
        "\n",
        "def train_cret(model, train_loader, num_epochs=2, device='cpu'):\n",
        "    print(f\"Training on device: {device}\")\n",
        "    model = model.to(device)\n",
        "    gees_loss = GEESLoss()\n",
        "    optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        model.train()\n",
        "        total_loss = 0\n",
        "        progress_bar = tqdm(train_loader, desc=f'Epoch {epoch+1}/{num_epochs}')\n",
        "\n",
        "        for batch_idx, (videos, text_ids, text_mask) in enumerate(progress_bar):\n",
        "            try:\n",
        "                videos = videos.to(device).float()\n",
        "                text_ids = text_ids.to(device)\n",
        "                text_mask = text_mask.to(device)\n",
        "\n",
        "                outputs = model(videos, text_ids, text_mask)\n",
        "\n",
        "                ccm_loss = F.cosine_embedding_loss(\n",
        "                    outputs['video_aligned'].view(-1, 768),\n",
        "                    outputs['text_aligned'].view(-1, 768),\n",
        "                    torch.ones(outputs['video_aligned'].size(0) * outputs['video_aligned'].size(1)).to(device)\n",
        "                )\n",
        "\n",
        "                gees_loss_val = gees_loss(outputs['video_global'], outputs['text_global'])\n",
        "                loss = ccm_loss + gees_loss_val\n",
        "\n",
        "                optimizer.zero_grad()\n",
        "                loss.backward()\n",
        "                optimizer.step()\n",
        "\n",
        "                total_loss += loss.item()\n",
        "                progress_bar.set_postfix({'loss': loss.item()})\n",
        "\n",
        "            except Exception as e:\n",
        "                print(f\"Error in batch {batch_idx}: {str(e)}\")\n",
        "                continue\n",
        "\n",
        "        avg_loss = total_loss / len(train_loader)\n",
        "        print(f'Epoch {epoch+1} completed, Average Loss: {avg_loss:.4f}')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nY39qLjLJmSt"
      },
      "source": [
        "## Preserve Trained Model"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 38,
      "metadata": {
        "id": "hVe9l_gmJqdp"
      },
      "outputs": [],
      "source": [
        "def save_checkpoint(model, optimizer, epoch, loss, filename):\n",
        "    torch.save({\n",
        "        'epoch': epoch,\n",
        "        'model_state_dict': model.state_dict(),\n",
        "        'optimizer_state_dict': optimizer.state_dict(),\n",
        "        'loss': loss,\n",
        "    }, filename)\n",
        "    print(f\"Checkpoint saved: {filename}\")\n",
        "\n",
        "def train_cret(model, train_loader, num_epochs=2, device='cpu'):\n",
        "    print(f\"Training on device: {device}\")\n",
        "    model = model.to(device)\n",
        "    gees_loss = GEESLoss()\n",
        "    optimizer = torch.optim.Adam(model.parameters(), lr=1e-4)\n",
        "\n",
        "    for epoch in range(num_epochs):\n",
        "        model.train()\n",
        "        total_loss = 0\n",
        "        progress_bar = tqdm(train_loader, desc=f'Epoch {epoch+1}/{num_epochs}')\n",
        "\n",
        "        for batch_idx, (videos, text_ids, text_mask) in enumerate(progress_bar):\n",
        "            try:\n",
        "                videos = videos.to(device).float()\n",
        "                text_ids = text_ids.to(device)\n",
        "                text_mask = text_mask.to(device)\n",
        "\n",
        "                outputs = model(videos, text_ids, text_mask)\n",
        "\n",
        "                ccm_loss = F.cosine_embedding_loss(\n",
        "                    outputs['video_aligned'].view(-1, 768),\n",
        "                    outputs['text_aligned'].view(-1, 768),\n",
        "                    torch.ones(outputs['video_aligned'].size(0) * outputs['video_aligned'].size(1)).to(device)\n",
        "                )\n",
        "\n",
        "                gees_loss_val = gees_loss(outputs['video_global'], outputs['text_global'])\n",
        "                loss = ccm_loss + gees_loss_val\n",
        "\n",
        "                optimizer.zero_grad()\n",
        "                loss.backward()\n",
        "                optimizer.step()\n",
        "\n",
        "                total_loss += loss.item()\n",
        "                progress_bar.set_postfix({'loss': loss.item()})\n",
        "\n",
        "            except Exception as e:\n",
        "                print(f\"Error in batch {batch_idx}: {str(e)}\")\n",
        "                continue\n",
        "\n",
        "        avg_loss = total_loss / len(train_loader)\n",
        "        print(f'Epoch {epoch+1} completed, Average Loss: {avg_loss:.4f}')\n",
        "\n",
        "        # Save checkpoint after each epoch\n",
        "        save_checkpoint(model, optimizer, epoch, avg_loss,\n",
        "                      f'cret_checkpoint_epoch_{epoch+1}.pt')"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "h44ldkPSJwr5"
      },
      "source": [
        "## Test Retrieval"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 39,
      "metadata": {
        "id": "TC3y3o4sJyGi"
      },
      "outputs": [],
      "source": [
        "def test_retrieval(model, video_path, query_text, tokenizer, device='cpu'):\n",
        "    model.eval()\n",
        "\n",
        "    # Load and process video\n",
        "    dataset = MSRVTTDataset([video_path], [query_text], tokenizer)\n",
        "    video, text_ids, text_mask = dataset[0]\n",
        "\n",
        "    with torch.no_grad():\n",
        "        video = video.unsqueeze(0).to(device)\n",
        "        text_ids = text_ids.unsqueeze(0).to(device)\n",
        "        text_mask = text_mask.unsqueeze(0).to(device)\n",
        "\n",
        "        outputs = model(video, text_ids, text_mask)\n",
        "\n",
        "        # Compute similarity score\n",
        "        video_emb = outputs['video_global']\n",
        "        text_emb = outputs['text_global']\n",
        "        similarity = F.cosine_similarity(video_emb, text_emb)\n",
        "\n",
        "        return similarity.item()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9wDi8JPgJ-t2"
      },
      "source": [
        "## Run Training"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": 40,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4H6VEo83J_6Q",
        "outputId": "0841e76d-903a-4660-a92a-9e336b313b88"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Found 999 video files\n",
            "Created test annotations file\n",
            "Prepared 999 video-caption pairs\n",
            "Training on device: cpu\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 1/2: 100%|██████████| 500/500 [1:22:54<00:00,  9.95s/it, loss=0.000117]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1 completed, Average Loss: 21.1536\n",
            "Checkpoint saved: cret_checkpoint_epoch_1.pt\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Epoch 2/2: 100%|██████████| 500/500 [1:25:22<00:00, 10.25s/it, loss=7.23e-5]\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 2 completed, Average Loss: 1.1550\n",
            "Checkpoint saved: cret_checkpoint_epoch_2.pt\n"
          ]
        }
      ],
      "source": [
        "def create_test_annotations():\n",
        "    # Get list of videos\n",
        "    video_files = sorted([f for f in os.listdir('MSRVTT') if f.endswith('.mp4')])\n",
        "    print(f\"Found {len(video_files)} video files\")\n",
        "\n",
        "    # Create annotations dictionary\n",
        "    annotations = {}\n",
        "    for video_file in video_files:\n",
        "        video_id = video_file[:-4]  # Remove .mp4\n",
        "        # Store multiple captions per video for better training\n",
        "        annotations[video_id] = [\n",
        "            {'caption': f'A video showing content from {video_id}'},\n",
        "            {'caption': f'This is video file {video_id}'},\n",
        "            {'caption': f'A clip from video sequence {video_id}'}\n",
        "        ]\n",
        "\n",
        "    # Save annotations\n",
        "    with open('MSRVTT/test_annotations.json', 'w') as f:\n",
        "        json.dump(annotations, f)\n",
        "\n",
        "    print(\"Created test annotations file\")\n",
        "    return annotations  # Added this line to return the annotations\n",
        "\n",
        "def main():\n",
        "    # Initialize tokenizer\n",
        "    tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')\n",
        "\n",
        "    # Set correct path\n",
        "    msrvtt_root = 'MSRVTT'\n",
        "    video_paths = sorted([os.path.join(msrvtt_root, f) for f in os.listdir(msrvtt_root) if f.endswith('.mp4')])\n",
        "    annotations = create_test_annotations()  # Now this will receive the returned annotations\n",
        "\n",
        "    # Get captions for each video\n",
        "    captions = []\n",
        "    for video_path in video_paths:\n",
        "        video_id = os.path.basename(video_path)[:-4]\n",
        "        if video_id in annotations:\n",
        "            caption = annotations[video_id][0]['caption']\n",
        "            captions.append(caption)\n",
        "\n",
        "    print(f\"Prepared {len(captions)} video-caption pairs\")\n",
        "    dataset = MSRVTTDataset(video_paths[:len(captions)], captions, tokenizer)\n",
        "    train_loader = DataLoader(dataset, batch_size=2, shuffle=True)\n",
        "\n",
        "    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "    model = CRET().to(device)\n",
        "\n",
        "    # Train model with progress bar\n",
        "    train_cret(model, train_loader, num_epochs=2, device=device)\n",
        "\n",
        "if __name__ == \"__main__\":\n",
        "    main()"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## CRET Model"
      ],
      "metadata": {
        "id": "DCXMwcww60c-"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Check what variables exist\n",
        "print(\"Current variables:\")\n",
        "%who"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xYaP7RcS9usW",
        "outputId": "80cfc834-ee5d-4f73-fc34-958378d44fbb"
      },
      "execution_count": 44,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Current variables:\n",
            "BertModel\t BertTokenizer\t CCMModule\t CRET\t DataLoader\t Dataset\t F\t GEESLoss\t List\t \n",
            "MSRVTTDataset\t Tuple\t VideoEncoder\t create_test_annotations\t cv2\t json\t main\t model_save_path\t models\t \n",
            "nn\t np\t os\t save_checkpoint\t test_dataset\t test_retrieval\t torch\t tqdm\t train_cret\t \n",
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Import required libraries\n",
        "import os\n",
        "import json\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "from transformers import BertTokenizer, BertModel\n",
        "import torchvision.models as models\n",
        "from typing import List, Tuple\n",
        "import numpy as np\n",
        "from tqdm import tqdm\n",
        "\n",
        "# 1. Define VideoEncoder\n",
        "class VideoEncoder(nn.Module):\n",
        "    def __init__(self, d_model=768):\n",
        "        super().__init__()\n",
        "        self.spatial_encoder = models.resnet50(pretrained=True)\n",
        "        self.spatial_encoder = nn.Sequential(*list(self.spatial_encoder.children())[:-1])\n",
        "        self.projection = nn.Linear(2048, d_model)\n",
        "        encoder_layer = nn.TransformerEncoderLayer(d_model=d_model, nhead=8)\n",
        "        self.temporal_encoder = nn.TransformerEncoder(encoder_layer, num_layers=3)\n",
        "\n",
        "    def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        B, F, C, H, W = x.shape\n",
        "        x = x.view(B*F, C, H, W)\n",
        "        spatial_features = self.spatial_encoder(x)\n",
        "        spatial_features = spatial_features.view(B, F, -1)\n",
        "        spatial_features = self.projection(spatial_features)\n",
        "        temporal_features = self.temporal_encoder(spatial_features.transpose(0, 1)).transpose(0, 1)\n",
        "        return temporal_features, spatial_features\n",
        "\n",
        "# 2. Define CCMModule\n",
        "class CCMModule(nn.Module):\n",
        "    def __init__(self, d_model=768, num_heads=8, num_queries=8):\n",
        "        super().__init__()\n",
        "        self.num_queries = num_queries\n",
        "        self.query_centers = nn.Parameter(torch.randn(num_queries, d_model))\n",
        "        self.decoder = nn.TransformerDecoderLayer(d_model=d_model, nhead=num_heads)\n",
        "\n",
        "    def forward(self, video_features: torch.Tensor, text_features: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        query = self.query_centers.unsqueeze(0).expand(video_features.size(0), -1, -1)\n",
        "        video_aligned = self.decoder(query, video_features)\n",
        "        text_aligned = self.decoder(query, text_features)\n",
        "        return video_aligned, text_aligned\n",
        "\n",
        "# 3. Define CRET\n",
        "class CRET(nn.Module):\n",
        "    def __init__(self, d_model=768):\n",
        "        super().__init__()\n",
        "        self.video_encoder = VideoEncoder(d_model)\n",
        "        self.text_encoder = BertModel.from_pretrained('bert-base-uncased')\n",
        "        self.ccm = CCMModule(d_model)\n",
        "\n",
        "    def forward(self, video: torch.Tensor, text_ids: torch.Tensor, text_mask: torch.Tensor):\n",
        "        temporal_features, spatial_features = self.video_encoder(video)\n",
        "        text_outputs = self.text_encoder(text_ids, attention_mask=text_mask)\n",
        "        text_features = text_outputs.last_hidden_state\n",
        "        video_aligned, text_aligned = self.ccm(spatial_features, text_features)\n",
        "        video_global = temporal_features.mean(1)\n",
        "        text_global = text_outputs.pooler_output\n",
        "        return {\n",
        "            'video_aligned': video_aligned,\n",
        "            'text_aligned': text_aligned,\n",
        "            'video_global': video_global,\n",
        "            'text_global': text_global\n",
        "        }\n",
        "\n",
        "# 4. Define GEESLoss\n",
        "class GEESLoss(nn.Module):\n",
        "    def __init__(self):\n",
        "        super().__init__()\n",
        "\n",
        "    def forward(self, video_features: torch.Tensor, text_features: torch.Tensor):\n",
        "        video_mean = video_features.mean(1)\n",
        "        video_cov = torch.bmm(video_features.transpose(1, 2), video_features) / video_features.size(1)\n",
        "        sim = torch.mm(text_features, video_mean.t())\n",
        "        cov_term = 0.5 * torch.bmm(torch.bmm(text_features.unsqueeze(1), video_cov), text_features.unsqueeze(-1)).squeeze()\n",
        "        logits = sim + cov_term\n",
        "        labels = torch.arange(video_features.size(0)).to(video_features.device)\n",
        "        return F.cross_entropy(logits, labels)\n",
        "\n",
        "# 5. Define MSRVTTDataset\n",
        "class MSRVTTDataset(Dataset):\n",
        "    def __init__(self, video_paths, captions, tokenizer):\n",
        "        self.video_paths = video_paths\n",
        "        self.captions = captions\n",
        "        self.tokenizer = tokenizer\n",
        "\n",
        "    def __len__(self):\n",
        "        return len(self.video_paths)\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        video = self.load_video(self.video_paths[idx])\n",
        "        caption = self.captions[idx]\n",
        "        tokens = self.tokenizer(caption, padding='max_length', truncation=True,\n",
        "                              max_length=128, return_tensors='pt')\n",
        "        return video, tokens.input_ids.squeeze(0), tokens.attention_mask.squeeze(0)\n",
        "\n",
        "    def load_video(self, path):\n",
        "        import cv2\n",
        "        frames = []\n",
        "        cap = cv2.VideoCapture(path)\n",
        "        total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
        "        indices = np.linspace(0, total_frames-1, 4, dtype=int)\n",
        "\n",
        "        for frame_idx in indices:\n",
        "            cap.set(cv2.CAP_PROP_POS_FRAMES, frame_idx)\n",
        "            ret, frame = cap.read()\n",
        "            if ret:\n",
        "                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)\n",
        "                frame = cv2.resize(frame, (224, 224))\n",
        "                frame = torch.from_numpy(frame).permute(2, 0, 1).float() / 255.0\n",
        "                frames.append(frame)\n",
        "\n",
        "        cap.release()\n",
        "        video_tensor = torch.stack(frames)\n",
        "        return video_tensor\n",
        "\n",
        "# Now initialize model and continue with your code\n",
        "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\n",
        "model = CRET().to(device)\n",
        "\n",
        "# Initialize tokenizer and continue with rest of your code..."
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xKtnWZGTF-nz",
        "outputId": "f63408df-8c01-4535-e716-5346ed697c4b"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
            "  warnings.warn(\n",
            "/usr/local/lib/python3.10/dist-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet50_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet50_Weights.DEFAULT` to get the most up-to-date weights.\n",
            "  warnings.warn(msg)\n",
            "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/transformer.py:379: UserWarning: enable_nested_tensor is True, but self.use_nested_tensor is False because encoder_layer.self_attn.batch_first was not True(use batch_first for better inference performance)\n",
            "  warnings.warn(\n",
            "/usr/local/lib/python3.10/dist-packages/huggingface_hub/utils/_auth.py:94: UserWarning: \n",
            "The secret `HF_TOKEN` does not exist in your Colab secrets.\n",
            "To authenticate with the Hugging Face Hub, create a token in your settings tab (https://huggingface.co/settings/tokens), set it as secret in your Google Colab and restart your session.\n",
            "You will be able to reuse this secret in all of your notebooks.\n",
            "Please note that authentication is recommended but still optional to access public models or datasets.\n",
            "  warnings.warn(\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Save the model\n",
        "model_save_path = 'cret_model.pth'\n",
        "torch.save(model.state_dict(), model_save_path)\n",
        "print(f\"Model saved to {model_save_path}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KNxV4kQzHC94",
        "outputId": "4f2deb9e-58fb-45b8-b3a4-fefe50aeef15"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model saved to cret_model.pth\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# 1. First update VideoEncoder\n",
        "class VideoEncoder(nn.Module):\n",
        "    def __init__(self, d_model=768):\n",
        "        super().__init__()\n",
        "        self.d_model = d_model\n",
        "        # Spatial encoder (ResNet)\n",
        "        self.spatial_encoder = models.resnet50(pretrained=True)\n",
        "        self.spatial_encoder = nn.Sequential(*list(self.spatial_encoder.children())[:-1])\n",
        "        self.projection = nn.Linear(2048, d_model)\n",
        "\n",
        "        # Temporal encoder with fixed parameters\n",
        "        encoder_layer = nn.TransformerEncoderLayer(\n",
        "            d_model=d_model,\n",
        "            nhead=8,\n",
        "            batch_first=True,\n",
        "            dim_feedforward=2048\n",
        "        )\n",
        "        self.temporal_encoder = nn.TransformerEncoder(encoder_layer, num_layers=3)\n",
        "\n",
        "    def forward(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        B, F, C, H, W = x.shape  # Batch, Frames, Channels, Height, Width\n",
        "\n",
        "        # Process each frame\n",
        "        x = x.view(B * F, C, H, W)\n",
        "        x = self.spatial_encoder(x)\n",
        "        x = x.squeeze(-1).squeeze(-1)  # Remove spatial dimensions\n",
        "        x = self.projection(x)\n",
        "\n",
        "        # Reshape for temporal processing\n",
        "        x = x.view(B, F, -1)  # [batch, frames, features]\n",
        "\n",
        "        # Apply temporal encoding\n",
        "        temporal_features = self.temporal_encoder(x)\n",
        "\n",
        "        return temporal_features, x\n",
        "\n",
        "# 2. Update CCMModule\n",
        "class CCMModule(nn.Module):\n",
        "    def __init__(self, d_model=768, num_heads=8, num_queries=8):\n",
        "        super().__init__()\n",
        "        self.num_queries = num_queries\n",
        "        self.d_model = d_model\n",
        "\n",
        "        # Learnable query centers\n",
        "        self.query_centers = nn.Parameter(torch.randn(1, num_queries, d_model))\n",
        "\n",
        "        # Transformer decoder layer with fixed parameters\n",
        "        self.decoder = nn.TransformerDecoderLayer(\n",
        "            d_model=d_model,\n",
        "            nhead=num_heads,\n",
        "            batch_first=True,\n",
        "            dim_feedforward=2048\n",
        "        )\n",
        "\n",
        "    def forward(self, video_features: torch.Tensor, text_features: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:\n",
        "        batch_size = video_features.size(0)\n",
        "\n",
        "        # Expand queries for batch\n",
        "        query = self.query_centers.expand(batch_size, -1, -1)\n",
        "\n",
        "        # Apply decoder to both modalities\n",
        "        video_aligned = self.decoder(query, video_features)\n",
        "        text_aligned = self.decoder(query, text_features)\n",
        "\n",
        "        return video_aligned, text_aligned\n",
        "\n",
        "# 3. Re-initialize the model with updated components\n",
        "model = CRET().to(device)"
      ],
      "metadata": {
        "id": "20fg-930HcsN"
      },
      "execution_count": 9,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def retrieve_video(query_text, top_k=5):\n",
        "    model.eval()\n",
        "    device = next(model.parameters()).device\n",
        "\n",
        "    # Process query\n",
        "    tokens = tokenizer(\n",
        "        query_text,\n",
        "        padding='max_length',\n",
        "        truncation=True,\n",
        "        max_length=128,\n",
        "        return_tensors='pt'\n",
        "    ).to(device)\n",
        "\n",
        "    # Get video paths (start with just a few videos)\n",
        "    video_paths = sorted([os.path.join('MSRVTT', f) for f in os.listdir('MSRVTT')\n",
        "                         if f.endswith('.mp4')])[:10]  # Start with 10 videos\n",
        "\n",
        "    results = []\n",
        "    with torch.no_grad():\n",
        "        for video_path in tqdm(video_paths, desc=\"Processing videos\"):\n",
        "            try:\n",
        "                # Load video\n",
        "                dataset = MSRVTTDataset([video_path], [\"dummy\"], tokenizer)\n",
        "                video, _, _ = dataset[0]\n",
        "                video = video.unsqueeze(0).to(device)\n",
        "\n",
        "                # Print shapes for debugging\n",
        "                if len(results) == 0:\n",
        "                    print(f\"Video shape: {video.shape}\")\n",
        "                    print(f\"Text IDs shape: {tokens.input_ids.shape}\")\n",
        "\n",
        "                # Get embeddings\n",
        "                outputs = model(video, tokens.input_ids, tokens.attention_mask)\n",
        "\n",
        "                # Calculate similarity\n",
        "                similarity = F.cosine_similarity(\n",
        "                    outputs['video_global'],\n",
        "                    outputs['text_global']\n",
        "                ).item()\n",
        "\n",
        "                results.append((video_path, similarity))\n",
        "\n",
        "            except Exception as e:\n",
        "                print(f\"Error processing {video_path}: {str(e)}\")\n",
        "                print(f\"Shapes at error:\")\n",
        "                print(f\"Video shape: {video.shape}\")\n",
        "                print(f\"Text shape: {tokens.input_ids.shape}\")\n",
        "                continue\n",
        "\n",
        "    # Sort by similarity\n",
        "    results.sort(key=lambda x: x[1], reverse=True)\n",
        "    return results[:top_k]\n",
        "\n",
        "# Test with a single query\n",
        "test_query = \"a person is cooking food\"\n",
        "print(f\"\\nQuery: {test_query}\")\n",
        "results = retrieve_video(test_query)\n",
        "print(\"\\nTop matching videos:\")\n",
        "for i, (video_path, similarity) in enumerate(results, 1):\n",
        "    print(f\"{i}. {os.path.basename(video_path)} (similarity: {similarity:.3f})\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "rscNwRFTHHrx",
        "outputId": "35046c32-dc73-43df-9f5a-6f5237ebc118"
      },
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Query: a person is cooking food\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\rProcessing videos:   0%|          | 0/10 [00:00<?, ?it/s]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Video shape: torch.Size([1, 4, 3, 224, 224])\n",
            "Text IDs shape: torch.Size([1, 128])\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "Processing videos: 100%|██████████| 10/10 [00:17<00:00,  1.75s/it]"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\n",
            "Top matching videos:\n",
            "1. video1000.mp4 (similarity: -0.011)\n",
            "2. video102.mp4 (similarity: -0.016)\n",
            "3. video0.mp4 (similarity: -0.033)\n",
            "4. video104.mp4 (similarity: -0.045)\n",
            "5. video1.mp4 (similarity: -0.050)\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Interface"
      ],
      "metadata": {
        "id": "i_YuiG8NIA7z"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install gradio"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uObOl6rFINuk",
        "outputId": "7263d118-d7d2-4c87-81df-c1426dfff83f"
      },
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: gradio in /usr/local/lib/python3.10/dist-packages (5.8.0)\n",
            "Requirement already satisfied: aiofiles<24.0,>=22.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (23.2.1)\n",
            "Requirement already satisfied: anyio<5.0,>=3.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (3.7.1)\n",
            "Requirement already satisfied: fastapi<1.0,>=0.115.2 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.115.6)\n",
            "Requirement already satisfied: ffmpy in /usr/local/lib/python3.10/dist-packages (from gradio) (0.4.0)\n",
            "Requirement already satisfied: gradio-client==1.5.1 in /usr/local/lib/python3.10/dist-packages (from gradio) (1.5.1)\n",
            "Requirement already satisfied: httpx>=0.24.1 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.28.0)\n",
            "Requirement already satisfied: huggingface-hub>=0.25.1 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.26.3)\n",
            "Requirement already satisfied: jinja2<4.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (3.1.4)\n",
            "Requirement already satisfied: markupsafe~=2.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (2.1.5)\n",
            "Requirement already satisfied: numpy<3.0,>=1.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (1.26.4)\n",
            "Requirement already satisfied: orjson~=3.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (3.10.12)\n",
            "Requirement already satisfied: packaging in /usr/local/lib/python3.10/dist-packages (from gradio) (24.2)\n",
            "Requirement already satisfied: pandas<3.0,>=1.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (2.2.2)\n",
            "Requirement already satisfied: pillow<12.0,>=8.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (11.0.0)\n",
            "Requirement already satisfied: pydantic>=2.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (2.10.3)\n",
            "Requirement already satisfied: pydub in /usr/local/lib/python3.10/dist-packages (from gradio) (0.25.1)\n",
            "Requirement already satisfied: python-multipart>=0.0.18 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.0.19)\n",
            "Requirement already satisfied: pyyaml<7.0,>=5.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (6.0.2)\n",
            "Requirement already satisfied: ruff>=0.2.2 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.8.2)\n",
            "Requirement already satisfied: safehttpx<0.2.0,>=0.1.6 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.1.6)\n",
            "Requirement already satisfied: semantic-version~=2.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (2.10.0)\n",
            "Requirement already satisfied: starlette<1.0,>=0.40.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.41.3)\n",
            "Requirement already satisfied: tomlkit<0.14.0,>=0.12.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.13.2)\n",
            "Requirement already satisfied: typer<1.0,>=0.12 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.15.0)\n",
            "Requirement already satisfied: typing-extensions~=4.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (4.12.2)\n",
            "Requirement already satisfied: uvicorn>=0.14.0 in /usr/local/lib/python3.10/dist-packages (from gradio) (0.32.1)\n",
            "Requirement already satisfied: fsspec in /usr/local/lib/python3.10/dist-packages (from gradio-client==1.5.1->gradio) (2024.10.0)\n",
            "Requirement already satisfied: websockets<15.0,>=10.0 in /usr/local/lib/python3.10/dist-packages (from gradio-client==1.5.1->gradio) (14.1)\n",
            "Requirement already satisfied: idna>=2.8 in /usr/local/lib/python3.10/dist-packages (from anyio<5.0,>=3.0->gradio) (3.10)\n",
            "Requirement already satisfied: sniffio>=1.1 in /usr/local/lib/python3.10/dist-packages (from anyio<5.0,>=3.0->gradio) (1.3.1)\n",
            "Requirement already satisfied: exceptiongroup in /usr/local/lib/python3.10/dist-packages (from anyio<5.0,>=3.0->gradio) (1.2.2)\n",
            "Requirement already satisfied: certifi in /usr/local/lib/python3.10/dist-packages (from httpx>=0.24.1->gradio) (2024.8.30)\n",
            "Requirement already satisfied: httpcore==1.* in /usr/local/lib/python3.10/dist-packages (from httpx>=0.24.1->gradio) (1.0.7)\n",
            "Requirement already satisfied: h11<0.15,>=0.13 in /usr/local/lib/python3.10/dist-packages (from httpcore==1.*->httpx>=0.24.1->gradio) (0.14.0)\n",
            "Requirement already satisfied: filelock in /usr/local/lib/python3.10/dist-packages (from huggingface-hub>=0.25.1->gradio) (3.16.1)\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.10/dist-packages (from huggingface-hub>=0.25.1->gradio) (2.32.3)\n",
            "Requirement already satisfied: tqdm>=4.42.1 in /usr/local/lib/python3.10/dist-packages (from huggingface-hub>=0.25.1->gradio) (4.66.6)\n",
            "Requirement already satisfied: python-dateutil>=2.8.2 in /usr/local/lib/python3.10/dist-packages (from pandas<3.0,>=1.0->gradio) (2.8.2)\n",
            "Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-packages (from pandas<3.0,>=1.0->gradio) (2024.2)\n",
            "Requirement already satisfied: tzdata>=2022.7 in /usr/local/lib/python3.10/dist-packages (from pandas<3.0,>=1.0->gradio) (2024.2)\n",
            "Requirement already satisfied: annotated-types>=0.6.0 in /usr/local/lib/python3.10/dist-packages (from pydantic>=2.0->gradio) (0.7.0)\n",
            "Requirement already satisfied: pydantic-core==2.27.1 in /usr/local/lib/python3.10/dist-packages (from pydantic>=2.0->gradio) (2.27.1)\n",
            "Requirement already satisfied: click>=8.0.0 in /usr/local/lib/python3.10/dist-packages (from typer<1.0,>=0.12->gradio) (8.1.7)\n",
            "Requirement already satisfied: shellingham>=1.3.0 in /usr/local/lib/python3.10/dist-packages (from typer<1.0,>=0.12->gradio) (1.5.4)\n",
            "Requirement already satisfied: rich>=10.11.0 in /usr/local/lib/python3.10/dist-packages (from typer<1.0,>=0.12->gradio) (13.9.4)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.8.2->pandas<3.0,>=1.0->gradio) (1.16.0)\n",
            "Requirement already satisfied: markdown-it-py>=2.2.0 in /usr/local/lib/python3.10/dist-packages (from rich>=10.11.0->typer<1.0,>=0.12->gradio) (3.0.0)\n",
            "Requirement already satisfied: pygments<3.0.0,>=2.13.0 in /usr/local/lib/python3.10/dist-packages (from rich>=10.11.0->typer<1.0,>=0.12->gradio) (2.18.0)\n",
            "Requirement already satisfied: charset-normalizer<4,>=2 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface-hub>=0.25.1->gradio) (3.4.0)\n",
            "Requirement already satisfied: urllib3<3,>=1.21.1 in /usr/local/lib/python3.10/dist-packages (from requests->huggingface-hub>=0.25.1->gradio) (2.2.3)\n",
            "Requirement already satisfied: mdurl~=0.1 in /usr/local/lib/python3.10/dist-packages (from markdown-it-py>=2.2.0->rich>=10.11.0->typer<1.0,>=0.12->gradio) (0.1.2)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "def retrieve_video_from_larger_set(query, num_videos=100, top_k=5):\n",
        "    \"\"\"\n",
        "    Retrieves the top-k most relevant video paths from a larger set of videos based on the given query.\n",
        "\n",
        "    Args:\n",
        "        query (str): The text query to search for.\n",
        "        num_videos (int): The number of videos to consider in the larger set.\n",
        "        top_k (int): The number of most relevant videos to return.\n",
        "\n",
        "    Returns:\n",
        "        list: A list of tuples, where each tuple contains the video path and the similarity score.\n",
        "    \"\"\"\n",
        "    # Implement the CRET model logic to retrieve the top-k most relevant video paths\n",
        "    # This might involve:\n",
        "    # 1. Encoding the query using a text encoder\n",
        "    # 2. Encoding the video metadata (titles, descriptions, etc.) using the same encoder\n",
        "    # 3. Calculating the similarity between the query and the video metadata\n",
        "    # 4. Sorting the videos by similarity and returning the top-k results\n",
        "\n",
        "    # For now, this is a placeholder implementation that simply returns a list of random video paths\n",
        "    import random\n",
        "    video_paths = [f\"video_{i}.mp4\" for i in range(num_videos)]\n",
        "    similarities = [random.uniform(0, 1) for _ in range(num_videos)]\n",
        "\n",
        "    # Sort the results by similarity in descending order\n",
        "    sorted_results = sorted(zip(video_paths, similarities), key=lambda x: x[1], reverse=True)\n",
        "\n",
        "    return sorted_results[:top_k]"
      ],
      "metadata": {
        "id": "CPX1psRoWwb_"
      },
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "import cv2\n",
        "from PIL import Image\n",
        "import numpy as np\n",
        "\n",
        "def process_video_for_display(video_path, max_frames=6):\n",
        "    try:\n",
        "        cap = cv2.VideoCapture(video_path)\n",
        "        if not cap.isOpened():\n",
        "            print(f\"Could not open video: {video_path}\")\n",
        "            return []\n",
        "\n",
        "        total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
        "        if total_frames == 0:\n",
        "            print(f\"Video {video_path} has no frames.\")\n",
        "            return []\n",
        "\n",
        "        indices = np.linspace(0, total_frames-1, max_frames, dtype=int)\n",
        "\n",
        "        frames = []\n",
        "        for idx in indices:\n",
        "            cap.set(cv2.CAP_PROP_POS_FRAMES, idx)\n",
        "            ret, frame = cap.read()\n",
        "            if ret:\n",
        "                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)\n",
        "                frames.append(Image.fromarray(frame))\n",
        "            else:\n",
        "                print(f\"Error reading frame {idx} from video {video_path}\")\n",
        "\n",
        "        cap.release()\n",
        "        return frames\n",
        "    except Exception as e:\n",
        "        print(f\"Error processing video {video_path}: {str(e)}\")\n",
        "        return []"
      ],
      "metadata": {
        "id": "95pyyM9mXE7c"
      },
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def video_search(query, num_results=5):\n",
        "    try:\n",
        "        # Process query through CRET model\n",
        "        results = retrieve_video_from_larger_set(query, num_videos=100, top_k=num_results)\n",
        "\n",
        "        if not results:\n",
        "            return [], \"No results found\"\n",
        "\n",
        "        # Process results for display\n",
        "        output_frames = []\n",
        "        captions = []\n",
        "\n",
        "        for video_path, similarity in results:\n",
        "            frames = process_video_for_display(video_path)\n",
        "            if frames:\n",
        "                output_frames.extend(frames)\n",
        "                video_name = os.path.basename(video_path)\n",
        "                captions.extend([f\"{video_name} (similarity: {similarity:.3f})\"] * len(frames))\n",
        "\n",
        "        if not output_frames:\n",
        "            return [], \"No frames could be extracted\"\n",
        "\n",
        "        return output_frames, f\"Found {len(results)} matching videos\"\n",
        "\n",
        "    except Exception as e:\n",
        "        return [], f\"Error: {str(e)}\""
      ],
      "metadata": {
        "id": "uKgIyQzKXOGb"
      },
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "from google.colab import files\n",
        "uploaded = files.upload()  # Follow the prompt to upload your video files"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "AzTc0IL4adZM",
        "outputId": "5f3d33b6-58d2-4d6c-a504-ef9f27900969"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "\n",
              "     <input type=\"file\" id=\"files-730ac05f-de3d-42af-b67e-b73d1192c52c\" name=\"files[]\" multiple disabled\n",
              "        style=\"border:none\" />\n",
              "     <output id=\"result-730ac05f-de3d-42af-b67e-b73d1192c52c\">\n",
              "      Upload widget is only available when the cell has been executed in the\n",
              "      current browser session. Please rerun this cell to enable.\n",
              "      </output>\n",
              "      <script>// Copyright 2017 Google LLC\n",
              "//\n",
              "// Licensed under the Apache License, Version 2.0 (the \"License\");\n",
              "// you may not use this file except in compliance with the License.\n",
              "// You may obtain a copy of the License at\n",
              "//\n",
              "//      http://www.apache.org/licenses/LICENSE-2.0\n",
              "//\n",
              "// Unless required by applicable law or agreed to in writing, software\n",
              "// distributed under the License is distributed on an \"AS IS\" BASIS,\n",
              "// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n",
              "// See the License for the specific language governing permissions and\n",
              "// limitations under the License.\n",
              "\n",
              "/**\n",
              " * @fileoverview Helpers for google.colab Python module.\n",
              " */\n",
              "(function(scope) {\n",
              "function span(text, styleAttributes = {}) {\n",
              "  const element = document.createElement('span');\n",
              "  element.textContent = text;\n",
              "  for (const key of Object.keys(styleAttributes)) {\n",
              "    element.style[key] = styleAttributes[key];\n",
              "  }\n",
              "  return element;\n",
              "}\n",
              "\n",
              "// Max number of bytes which will be uploaded at a time.\n",
              "const MAX_PAYLOAD_SIZE = 100 * 1024;\n",
              "\n",
              "function _uploadFiles(inputId, outputId) {\n",
              "  const steps = uploadFilesStep(inputId, outputId);\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  // Cache steps on the outputElement to make it available for the next call\n",
              "  // to uploadFilesContinue from Python.\n",
              "  outputElement.steps = steps;\n",
              "\n",
              "  return _uploadFilesContinue(outputId);\n",
              "}\n",
              "\n",
              "// This is roughly an async generator (not supported in the browser yet),\n",
              "// where there are multiple asynchronous steps and the Python side is going\n",
              "// to poll for completion of each step.\n",
              "// This uses a Promise to block the python side on completion of each step,\n",
              "// then passes the result of the previous step as the input to the next step.\n",
              "function _uploadFilesContinue(outputId) {\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  const steps = outputElement.steps;\n",
              "\n",
              "  const next = steps.next(outputElement.lastPromiseValue);\n",
              "  return Promise.resolve(next.value.promise).then((value) => {\n",
              "    // Cache the last promise value to make it available to the next\n",
              "    // step of the generator.\n",
              "    outputElement.lastPromiseValue = value;\n",
              "    return next.value.response;\n",
              "  });\n",
              "}\n",
              "\n",
              "/**\n",
              " * Generator function which is called between each async step of the upload\n",
              " * process.\n",
              " * @param {string} inputId Element ID of the input file picker element.\n",
              " * @param {string} outputId Element ID of the output display.\n",
              " * @return {!Iterable<!Object>} Iterable of next steps.\n",
              " */\n",
              "function* uploadFilesStep(inputId, outputId) {\n",
              "  const inputElement = document.getElementById(inputId);\n",
              "  inputElement.disabled = false;\n",
              "\n",
              "  const outputElement = document.getElementById(outputId);\n",
              "  outputElement.innerHTML = '';\n",
              "\n",
              "  const pickedPromise = new Promise((resolve) => {\n",
              "    inputElement.addEventListener('change', (e) => {\n",
              "      resolve(e.target.files);\n",
              "    });\n",
              "  });\n",
              "\n",
              "  const cancel = document.createElement('button');\n",
              "  inputElement.parentElement.appendChild(cancel);\n",
              "  cancel.textContent = 'Cancel upload';\n",
              "  const cancelPromise = new Promise((resolve) => {\n",
              "    cancel.onclick = () => {\n",
              "      resolve(null);\n",
              "    };\n",
              "  });\n",
              "\n",
              "  // Wait for the user to pick the files.\n",
              "  const files = yield {\n",
              "    promise: Promise.race([pickedPromise, cancelPromise]),\n",
              "    response: {\n",
              "      action: 'starting',\n",
              "    }\n",
              "  };\n",
              "\n",
              "  cancel.remove();\n",
              "\n",
              "  // Disable the input element since further picks are not allowed.\n",
              "  inputElement.disabled = true;\n",
              "\n",
              "  if (!files) {\n",
              "    return {\n",
              "      response: {\n",
              "        action: 'complete',\n",
              "      }\n",
              "    };\n",
              "  }\n",
              "\n",
              "  for (const file of files) {\n",
              "    const li = document.createElement('li');\n",
              "    li.append(span(file.name, {fontWeight: 'bold'}));\n",
              "    li.append(span(\n",
              "        `(${file.type || 'n/a'}) - ${file.size} bytes, ` +\n",
              "        `last modified: ${\n",
              "            file.lastModifiedDate ? file.lastModifiedDate.toLocaleDateString() :\n",
              "                                    'n/a'} - `));\n",
              "    const percent = span('0% done');\n",
              "    li.appendChild(percent);\n",
              "\n",
              "    outputElement.appendChild(li);\n",
              "\n",
              "    const fileDataPromise = new Promise((resolve) => {\n",
              "      const reader = new FileReader();\n",
              "      reader.onload = (e) => {\n",
              "        resolve(e.target.result);\n",
              "      };\n",
              "      reader.readAsArrayBuffer(file);\n",
              "    });\n",
              "    // Wait for the data to be ready.\n",
              "    let fileData = yield {\n",
              "      promise: fileDataPromise,\n",
              "      response: {\n",
              "        action: 'continue',\n",
              "      }\n",
              "    };\n",
              "\n",
              "    // Use a chunked sending to avoid message size limits. See b/62115660.\n",
              "    let position = 0;\n",
              "    do {\n",
              "      const length = Math.min(fileData.byteLength - position, MAX_PAYLOAD_SIZE);\n",
              "      const chunk = new Uint8Array(fileData, position, length);\n",
              "      position += length;\n",
              "\n",
              "      const base64 = btoa(String.fromCharCode.apply(null, chunk));\n",
              "      yield {\n",
              "        response: {\n",
              "          action: 'append',\n",
              "          file: file.name,\n",
              "          data: base64,\n",
              "        },\n",
              "      };\n",
              "\n",
              "      let percentDone = fileData.byteLength === 0 ?\n",
              "          100 :\n",
              "          Math.round((position / fileData.byteLength) * 100);\n",
              "      percent.textContent = `${percentDone}% done`;\n",
              "\n",
              "    } while (position < fileData.byteLength);\n",
              "  }\n",
              "\n",
              "  // All done.\n",
              "  yield {\n",
              "    response: {\n",
              "      action: 'complete',\n",
              "    }\n",
              "  };\n",
              "}\n",
              "\n",
              "scope.google = scope.google || {};\n",
              "scope.google.colab = scope.google.colab || {};\n",
              "scope.google.colab._files = {\n",
              "  _uploadFiles,\n",
              "  _uploadFilesContinue,\n",
              "};\n",
              "})(self);\n",
              "</script> "
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Saving video0.mp4 to video0.mp4\n",
            "Saving video1.mp4 to video1.mp4\n",
            "Saving video2.mp4 to video2.mp4\n",
            "Saving video3.mp4 to video3.mp4\n",
            "Saving video4.mp4 to video4.mp4\n",
            "Saving video5.mp4 to video5.mp4\n",
            "Saving video6.mp4 to video6.mp4\n",
            "Saving video7.mp4 to video7.mp4\n",
            "Saving video8.mp4 to video8.mp4\n",
            "Saving video9.mp4 to video9.mp4\n",
            "Saving video10.mp4 to video10.mp4\n",
            "Saving video11.mp4 to video11.mp4\n",
            "Saving video12.mp4 to video12.mp4\n",
            "Saving video13.mp4 to video13.mp4\n",
            "Saving video14.mp4 to video14.mp4\n",
            "Saving video15.mp4 to video15.mp4\n",
            "Saving video16.mp4 to video16.mp4\n",
            "Saving video17.mp4 to video17.mp4\n",
            "Saving video18.mp4 to video18.mp4\n",
            "Saving video19.mp4 to video19.mp4\n",
            "Saving video20.mp4 to video20.mp4\n",
            "Saving video21.mp4 to video21.mp4\n",
            "Saving video22.mp4 to video22.mp4\n",
            "Saving video23.mp4 to video23.mp4\n",
            "Saving video24.mp4 to video24.mp4\n",
            "Saving video25.mp4 to video25.mp4\n",
            "Saving video26.mp4 to video26.mp4\n",
            "Saving video27.mp4 to video27.mp4\n",
            "Saving video28.mp4 to video28.mp4\n",
            "Saving video29.mp4 to video29.mp4\n",
            "Saving video30.mp4 to video30.mp4\n",
            "Saving video31.mp4 to video31.mp4\n",
            "Saving video32.mp4 to video32.mp4\n",
            "Saving video33.mp4 to video33.mp4\n",
            "Saving video34.mp4 to video34.mp4\n",
            "Saving video35.mp4 to video35.mp4\n",
            "Saving video36.mp4 to video36.mp4\n",
            "Saving video37.mp4 to video37.mp4\n",
            "Saving video38.mp4 to video38.mp4\n",
            "Saving video39.mp4 to video39.mp4\n",
            "Saving video40.mp4 to video40.mp4\n",
            "Saving video41.mp4 to video41.mp4\n",
            "Saving video42.mp4 to video42.mp4\n",
            "Saving video43.mp4 to video43.mp4\n",
            "Saving video44.mp4 to video44.mp4\n",
            "Saving video45.mp4 to video45.mp4\n",
            "Saving video46.mp4 to video46.mp4\n",
            "Saving video47.mp4 to video47.mp4\n",
            "Saving video48.mp4 to video48.mp4\n",
            "Saving video49.mp4 to video49.mp4\n",
            "Saving video50.mp4 to video50.mp4\n",
            "Saving video51.mp4 to video51.mp4\n",
            "Saving video52.mp4 to video52.mp4\n",
            "Saving video53.mp4 to video53.mp4\n",
            "Saving video54.mp4 to video54.mp4\n",
            "Saving video55.mp4 to video55.mp4\n",
            "Saving video56.mp4 to video56.mp4\n",
            "Saving video57.mp4 to video57.mp4\n",
            "Saving video58.mp4 to video58.mp4\n",
            "Saving video59.mp4 to video59.mp4\n",
            "Saving video60.mp4 to video60.mp4\n",
            "Saving video61.mp4 to video61.mp4\n",
            "Saving video62.mp4 to video62.mp4\n",
            "Saving video63.mp4 to video63.mp4\n",
            "Saving video64.mp4 to video64.mp4\n",
            "Saving video65.mp4 to video65.mp4\n",
            "Saving video66.mp4 to video66.mp4\n",
            "Saving video67.mp4 to video67.mp4\n",
            "Saving video68.mp4 to video68.mp4\n",
            "Saving video69.mp4 to video69.mp4\n",
            "Saving video70.mp4 to video70.mp4\n",
            "Saving video71.mp4 to video71.mp4\n",
            "Saving video72.mp4 to video72.mp4\n",
            "Saving video73.mp4 to video73.mp4\n",
            "Saving video74.mp4 to video74.mp4\n",
            "Saving video75.mp4 to video75.mp4\n",
            "Saving video76.mp4 to video76.mp4\n",
            "Saving video77.mp4 to video77.mp4\n",
            "Saving video78.mp4 to video78.mp4\n",
            "Saving video79.mp4 to video79.mp4\n",
            "Saving video80.mp4 to video80.mp4\n",
            "Saving video81.mp4 to video81.mp4\n",
            "Saving video82.mp4 to video82.mp4\n",
            "Saving video83.mp4 to video83.mp4\n",
            "Saving video84.mp4 to video84.mp4\n",
            "Saving video85.mp4 to video85.mp4\n",
            "Saving video86.mp4 to video86.mp4\n",
            "Saving video87.mp4 to video87.mp4\n",
            "Saving video88.mp4 to video88.mp4\n",
            "Saving video89.mp4 to video89.mp4\n",
            "Saving video90.mp4 to video90.mp4\n",
            "Saving video91.mp4 to video91.mp4\n",
            "Saving video92.mp4 to video92.mp4\n",
            "Saving video93.mp4 to video93.mp4\n",
            "Saving video94.mp4 to video94.mp4\n",
            "Saving video95.mp4 to video95.mp4\n",
            "Saving video96.mp4 to video96.mp4\n",
            "Saving video97.mp4 to video97.mp4\n",
            "Saving video98.mp4 to video98.mp4\n",
            "Saving video99.mp4 to video99.mp4\n",
            "Saving video100.mp4 to video100.mp4\n",
            "Saving video101.mp4 to video101.mp4\n",
            "Saving video102.mp4 to video102.mp4\n",
            "Saving video103.mp4 to video103.mp4\n",
            "Saving video104.mp4 to video104.mp4\n",
            "Saving video105.mp4 to video105.mp4\n",
            "Saving video106.mp4 to video106.mp4\n",
            "Saving video107.mp4 to video107.mp4\n",
            "Saving video108.mp4 to video108.mp4\n",
            "Saving video109.mp4 to video109.mp4\n",
            "Saving video110.mp4 to video110.mp4\n",
            "Saving video111.mp4 to video111.mp4\n",
            "Saving video112.mp4 to video112.mp4\n",
            "Saving video113.mp4 to video113.mp4\n",
            "Saving video114.mp4 to video114.mp4\n",
            "Saving video115.mp4 to video115.mp4\n",
            "Saving video116.mp4 to video116.mp4\n",
            "Saving video117.mp4 to video117.mp4\n",
            "Saving video118.mp4 to video118.mp4\n",
            "Saving video119.mp4 to video119.mp4\n",
            "Saving video120.mp4 to video120.mp4\n",
            "Saving video121.mp4 to video121.mp4\n",
            "Saving video122.mp4 to video122.mp4\n",
            "Saving video123.mp4 to video123.mp4\n",
            "Saving video124.mp4 to video124.mp4\n",
            "Saving video125.mp4 to video125.mp4\n",
            "Saving video126.mp4 to video126.mp4\n",
            "Saving video127.mp4 to video127.mp4\n",
            "Saving video128.mp4 to video128.mp4\n",
            "Saving video129.mp4 to video129.mp4\n",
            "Saving video130.mp4 to video130.mp4\n",
            "Saving video131.mp4 to video131.mp4\n",
            "Saving video132.mp4 to video132.mp4\n",
            "Saving video133.mp4 to video133.mp4\n",
            "Saving video134.mp4 to video134.mp4\n",
            "Saving video135.mp4 to video135.mp4\n",
            "Saving video136.mp4 to video136.mp4\n",
            "Saving video137.mp4 to video137.mp4\n",
            "Saving video138.mp4 to video138.mp4\n",
            "Saving video139.mp4 to video139.mp4\n",
            "Saving video140.mp4 to video140.mp4\n",
            "Saving video141.mp4 to video141.mp4\n",
            "Saving video142.mp4 to video142.mp4\n",
            "Saving video143.mp4 to video143.mp4\n",
            "Saving video144.mp4 to video144.mp4\n",
            "Saving video145.mp4 to video145.mp4\n",
            "Saving video146.mp4 to video146.mp4\n",
            "Saving video147.mp4 to video147.mp4\n",
            "Saving video148.mp4 to video148.mp4\n",
            "Saving video149.mp4 to video149.mp4\n",
            "Saving video150.mp4 to video150.mp4\n",
            "Saving video151.mp4 to video151.mp4\n",
            "Saving video152.mp4 to video152.mp4\n",
            "Saving video153.mp4 to video153.mp4\n",
            "Saving video154.mp4 to video154.mp4\n",
            "Saving video155.mp4 to video155.mp4\n",
            "Saving video156.mp4 to video156.mp4\n",
            "Saving video157.mp4 to video157.mp4\n",
            "Saving video158.mp4 to video158.mp4\n",
            "Saving video159.mp4 to video159.mp4\n",
            "Saving video160.mp4 to video160.mp4\n",
            "Saving video161.mp4 to video161.mp4\n",
            "Saving video162.mp4 to video162.mp4\n",
            "Saving video163.mp4 to video163.mp4\n",
            "Saving video164.mp4 to video164.mp4\n",
            "Saving video165.mp4 to video165.mp4\n",
            "Saving video166.mp4 to video166.mp4\n",
            "Saving video167.mp4 to video167.mp4\n",
            "Saving video168.mp4 to video168.mp4\n",
            "Saving video169.mp4 to video169.mp4\n",
            "Saving video170.mp4 to video170.mp4\n",
            "Saving video171.mp4 to video171.mp4\n",
            "Saving video172.mp4 to video172.mp4\n",
            "Saving video173.mp4 to video173.mp4\n",
            "Saving video174.mp4 to video174.mp4\n",
            "Saving video175.mp4 to video175.mp4\n",
            "Saving video176.mp4 to video176.mp4\n",
            "Saving video177.mp4 to video177.mp4\n",
            "Saving video178.mp4 to video178.mp4\n",
            "Saving video179.mp4 to video179.mp4\n",
            "Saving video180.mp4 to video180.mp4\n",
            "Saving video181.mp4 to video181.mp4\n",
            "Saving video182.mp4 to video182.mp4\n",
            "Saving video183.mp4 to video183.mp4\n",
            "Saving video184.mp4 to video184.mp4\n",
            "Saving video185.mp4 to video185.mp4\n",
            "Saving video186.mp4 to video186.mp4\n",
            "Saving video187.mp4 to video187.mp4\n",
            "Saving video188.mp4 to video188.mp4\n",
            "Saving video189.mp4 to video189.mp4\n",
            "Saving video190.mp4 to video190.mp4\n",
            "Saving video191.mp4 to video191.mp4\n",
            "Saving video192.mp4 to video192.mp4\n",
            "Saving video193.mp4 to video193.mp4\n",
            "Saving video194.mp4 to video194.mp4\n",
            "Saving video195.mp4 to video195.mp4\n",
            "Saving video196.mp4 to video196.mp4\n",
            "Saving video197.mp4 to video197.mp4\n",
            "Saving video198.mp4 to video198.mp4\n",
            "Saving video199.mp4 to video199.mp4\n",
            "Saving video200.mp4 to video200.mp4\n",
            "Saving video201.mp4 to video201.mp4\n",
            "Saving video202.mp4 to video202.mp4\n",
            "Saving video203.mp4 to video203.mp4\n",
            "Saving video204.mp4 to video204.mp4\n",
            "Saving video205.mp4 to video205.mp4\n",
            "Saving video206.mp4 to video206.mp4\n",
            "Saving video207.mp4 to video207.mp4\n",
            "Saving video208.mp4 to video208.mp4\n",
            "Saving video209.mp4 to video209.mp4\n",
            "Saving video210.mp4 to video210.mp4\n",
            "Saving video211.mp4 to video211.mp4\n",
            "Saving video212.mp4 to video212.mp4\n",
            "Saving video213.mp4 to video213.mp4\n",
            "Saving video214.mp4 to video214.mp4\n",
            "Saving video215.mp4 to video215.mp4\n",
            "Saving video216.mp4 to video216.mp4\n",
            "Saving video217.mp4 to video217.mp4\n",
            "Saving video218.mp4 to video218.mp4\n",
            "Saving video219.mp4 to video219.mp4\n",
            "Saving video220.mp4 to video220.mp4\n",
            "Saving video221.mp4 to video221.mp4\n",
            "Saving video222.mp4 to video222.mp4\n",
            "Saving video223.mp4 to video223.mp4\n",
            "Saving video224.mp4 to video224.mp4\n",
            "Saving video225.mp4 to video225.mp4\n",
            "Saving video226.mp4 to video226.mp4\n",
            "Saving video227.mp4 to video227.mp4\n",
            "Saving video228.mp4 to video228.mp4\n",
            "Saving video229.mp4 to video229.mp4\n",
            "Saving video230.mp4 to video230.mp4\n",
            "Saving video231.mp4 to video231.mp4\n",
            "Saving video232.mp4 to video232.mp4\n",
            "Saving video233.mp4 to video233.mp4\n",
            "Saving video234.mp4 to video234.mp4\n",
            "Saving video235.mp4 to video235.mp4\n",
            "Saving video236.mp4 to video236.mp4\n",
            "Saving video237.mp4 to video237.mp4\n",
            "Saving video238.mp4 to video238.mp4\n",
            "Saving video239.mp4 to video239.mp4\n",
            "Saving video240.mp4 to video240.mp4\n",
            "Saving video241.mp4 to video241.mp4\n",
            "Saving video242.mp4 to video242.mp4\n",
            "Saving video243.mp4 to video243.mp4\n",
            "Saving video244.mp4 to video244.mp4\n",
            "Saving video245.mp4 to video245.mp4\n",
            "Saving video246.mp4 to video246.mp4\n",
            "Saving video247.mp4 to video247.mp4\n",
            "Saving video248.mp4 to video248.mp4\n",
            "Saving video249.mp4 to video249.mp4\n",
            "Saving video250.mp4 to video250.mp4\n",
            "Saving video251.mp4 to video251.mp4\n",
            "Saving video252.mp4 to video252.mp4\n",
            "Saving video253.mp4 to video253.mp4\n",
            "Saving video254.mp4 to video254.mp4\n",
            "Saving video255.mp4 to video255.mp4\n",
            "Saving video256.mp4 to video256.mp4\n",
            "Saving video257.mp4 to video257.mp4\n",
            "Saving video258.mp4 to video258.mp4\n",
            "Saving video259.mp4 to video259.mp4\n",
            "Saving video260.mp4 to video260.mp4\n",
            "Saving video261.mp4 to video261.mp4\n",
            "Saving video262.mp4 to video262.mp4\n",
            "Saving video263.mp4 to video263.mp4\n",
            "Saving video264.mp4 to video264.mp4\n",
            "Saving video265.mp4 to video265.mp4\n",
            "Saving video266.mp4 to video266.mp4\n",
            "Saving video267.mp4 to video267.mp4\n",
            "Saving video268.mp4 to video268.mp4\n",
            "Saving video269.mp4 to video269.mp4\n",
            "Saving video270.mp4 to video270.mp4\n",
            "Saving video271.mp4 to video271.mp4\n",
            "Saving video272.mp4 to video272.mp4\n",
            "Saving video273.mp4 to video273.mp4\n",
            "Saving video274.mp4 to video274.mp4\n",
            "Saving video275.mp4 to video275.mp4\n",
            "Saving video276.mp4 to video276.mp4\n",
            "Saving video277.mp4 to video277.mp4\n",
            "Saving video278.mp4 to video278.mp4\n",
            "Saving video279.mp4 to video279.mp4\n",
            "Saving video280.mp4 to video280.mp4\n",
            "Saving video281.mp4 to video281.mp4\n",
            "Saving video282.mp4 to video282.mp4\n",
            "Saving video283.mp4 to video283.mp4\n",
            "Saving video284.mp4 to video284.mp4\n",
            "Saving video285.mp4 to video285.mp4\n",
            "Saving video286.mp4 to video286.mp4\n",
            "Saving video287.mp4 to video287.mp4\n",
            "Saving video288.mp4 to video288.mp4\n",
            "Saving video289.mp4 to video289.mp4\n",
            "Saving video290.mp4 to video290.mp4\n",
            "Saving video291.mp4 to video291.mp4\n",
            "Saving video292.mp4 to video292.mp4\n",
            "Saving video293.mp4 to video293.mp4\n",
            "Saving video294.mp4 to video294.mp4\n",
            "Saving video295.mp4 to video295.mp4\n",
            "Saving video296.mp4 to video296.mp4\n",
            "Saving video297.mp4 to video297.mp4\n",
            "Saving video298.mp4 to video298.mp4\n",
            "Saving video299.mp4 to video299.mp4\n",
            "Saving video300.mp4 to video300.mp4\n",
            "Saving video301.mp4 to video301.mp4\n",
            "Saving video302.mp4 to video302.mp4\n",
            "Saving video303.mp4 to video303.mp4\n",
            "Saving video304.mp4 to video304.mp4\n",
            "Saving video305.mp4 to video305.mp4\n",
            "Saving video306.mp4 to video306.mp4\n",
            "Saving video307.mp4 to video307.mp4\n",
            "Saving video308.mp4 to video308.mp4\n",
            "Saving video309.mp4 to video309.mp4\n",
            "Saving video310.mp4 to video310.mp4\n",
            "Saving video311.mp4 to video311.mp4\n",
            "Saving video312.mp4 to video312.mp4\n",
            "Saving video313.mp4 to video313.mp4\n",
            "Saving video314.mp4 to video314.mp4\n",
            "Saving video315.mp4 to video315.mp4\n",
            "Saving video316.mp4 to video316.mp4\n",
            "Saving video317.mp4 to video317.mp4\n",
            "Saving video318.mp4 to video318.mp4\n",
            "Saving video319.mp4 to video319.mp4\n",
            "Saving video320.mp4 to video320.mp4\n",
            "Saving video321.mp4 to video321.mp4\n",
            "Saving video322.mp4 to video322.mp4\n",
            "Saving video323.mp4 to video323.mp4\n",
            "Saving video324.mp4 to video324.mp4\n",
            "Saving video325.mp4 to video325.mp4\n",
            "Saving video326.mp4 to video326.mp4\n",
            "Saving video327.mp4 to video327.mp4\n",
            "Saving video328.mp4 to video328.mp4\n",
            "Saving video329.mp4 to video329.mp4\n",
            "Saving video330.mp4 to video330.mp4\n",
            "Saving video331.mp4 to video331.mp4\n",
            "Saving video332.mp4 to video332.mp4\n",
            "Saving video333.mp4 to video333.mp4\n",
            "Saving video334.mp4 to video334.mp4\n",
            "Saving video335.mp4 to video335.mp4\n",
            "Saving video336.mp4 to video336.mp4\n",
            "Saving video337.mp4 to video337.mp4\n",
            "Saving video338.mp4 to video338.mp4\n",
            "Saving video339.mp4 to video339.mp4\n",
            "Saving video340.mp4 to video340.mp4\n",
            "Saving video341.mp4 to video341.mp4\n",
            "Saving video342.mp4 to video342.mp4\n",
            "Saving video343.mp4 to video343.mp4\n",
            "Saving video344.mp4 to video344.mp4\n",
            "Saving video345.mp4 to video345.mp4\n",
            "Saving video346.mp4 to video346.mp4\n",
            "Saving video347.mp4 to video347.mp4\n",
            "Saving video348.mp4 to video348.mp4\n",
            "Saving video349.mp4 to video349.mp4\n",
            "Saving video350.mp4 to video350.mp4\n",
            "Saving video351.mp4 to video351.mp4\n",
            "Saving video352.mp4 to video352.mp4\n",
            "Saving video353.mp4 to video353.mp4\n",
            "Saving video354.mp4 to video354.mp4\n",
            "Saving video355.mp4 to video355.mp4\n",
            "Saving video356.mp4 to video356.mp4\n",
            "Saving video357.mp4 to video357.mp4\n",
            "Saving video358.mp4 to video358.mp4\n",
            "Saving video359.mp4 to video359.mp4\n",
            "Saving video360.mp4 to video360.mp4\n",
            "Saving video361.mp4 to video361.mp4\n",
            "Saving video362.mp4 to video362.mp4\n",
            "Saving video363.mp4 to video363.mp4\n",
            "Saving video364.mp4 to video364.mp4\n",
            "Saving video365.mp4 to video365.mp4\n",
            "Saving video366.mp4 to video366.mp4\n",
            "Saving video367.mp4 to video367.mp4\n",
            "Saving video368.mp4 to video368.mp4\n",
            "Saving video369.mp4 to video369.mp4\n",
            "Saving video370.mp4 to video370.mp4\n",
            "Saving video371.mp4 to video371.mp4\n",
            "Saving video372.mp4 to video372.mp4\n",
            "Saving video373.mp4 to video373.mp4\n",
            "Saving video374.mp4 to video374.mp4\n",
            "Saving video375.mp4 to video375.mp4\n",
            "Saving video376.mp4 to video376.mp4\n",
            "Saving video377.mp4 to video377.mp4\n",
            "Saving video378.mp4 to video378.mp4\n",
            "Saving video379.mp4 to video379.mp4\n",
            "Saving video380.mp4 to video380.mp4\n",
            "Saving video381.mp4 to video381.mp4\n",
            "Saving video382.mp4 to video382.mp4\n",
            "Saving video383.mp4 to video383.mp4\n",
            "Saving video384.mp4 to video384.mp4\n",
            "Saving video385.mp4 to video385.mp4\n",
            "Saving video386.mp4 to video386.mp4\n",
            "Saving video387.mp4 to video387.mp4\n",
            "Saving video388.mp4 to video388.mp4\n",
            "Saving video389.mp4 to video389.mp4\n",
            "Saving video390.mp4 to video390.mp4\n",
            "Saving video391.mp4 to video391.mp4\n",
            "Saving video392.mp4 to video392.mp4\n",
            "Saving video393.mp4 to video393.mp4\n",
            "Saving video394.mp4 to video394.mp4\n",
            "Saving video395.mp4 to video395.mp4\n",
            "Saving video396.mp4 to video396.mp4\n",
            "Saving video397.mp4 to video397.mp4\n",
            "Saving video398.mp4 to video398.mp4\n",
            "Saving video399.mp4 to video399.mp4\n",
            "Saving video400.mp4 to video400.mp4\n",
            "Saving video401.mp4 to video401.mp4\n",
            "Saving video402.mp4 to video402.mp4\n",
            "Saving video403.mp4 to video403.mp4\n",
            "Saving video404.mp4 to video404.mp4\n",
            "Saving video405.mp4 to video405.mp4\n",
            "Saving video406.mp4 to video406.mp4\n",
            "Saving video407.mp4 to video407.mp4\n",
            "Saving video408.mp4 to video408.mp4\n",
            "Saving video409.mp4 to video409.mp4\n",
            "Saving video410.mp4 to video410.mp4\n",
            "Saving video411.mp4 to video411.mp4\n",
            "Saving video412.mp4 to video412.mp4\n",
            "Saving video413.mp4 to video413.mp4\n",
            "Saving video414.mp4 to video414.mp4\n",
            "Saving video415.mp4 to video415.mp4\n",
            "Saving video416.mp4 to video416.mp4\n",
            "Saving video417.mp4 to video417.mp4\n",
            "Saving video418.mp4 to video418.mp4\n",
            "Saving video419.mp4 to video419.mp4\n",
            "Saving video420.mp4 to video420.mp4\n",
            "Saving video421.mp4 to video421.mp4\n",
            "Saving video422.mp4 to video422.mp4\n",
            "Saving video423.mp4 to video423.mp4\n",
            "Saving video424.mp4 to video424.mp4\n",
            "Saving video425.mp4 to video425.mp4\n",
            "Saving video426.mp4 to video426.mp4\n",
            "Saving video427.mp4 to video427.mp4\n",
            "Saving video428.mp4 to video428.mp4\n",
            "Saving video429.mp4 to video429.mp4\n",
            "Saving video430.mp4 to video430.mp4\n",
            "Saving video431.mp4 to video431.mp4\n",
            "Saving video432.mp4 to video432.mp4\n",
            "Saving video433.mp4 to video433.mp4\n",
            "Saving video434.mp4 to video434.mp4\n",
            "Saving video435.mp4 to video435.mp4\n",
            "Saving video436.mp4 to video436.mp4\n",
            "Saving video437.mp4 to video437.mp4\n",
            "Saving video438.mp4 to video438.mp4\n",
            "Saving video439.mp4 to video439.mp4\n",
            "Saving video440.mp4 to video440.mp4\n",
            "Saving video441.mp4 to video441.mp4\n",
            "Saving video442.mp4 to video442.mp4\n",
            "Saving video443.mp4 to video443.mp4\n",
            "Saving video444.mp4 to video444.mp4\n",
            "Saving video445.mp4 to video445.mp4\n",
            "Saving video446.mp4 to video446.mp4\n",
            "Saving video447.mp4 to video447.mp4\n",
            "Saving video448.mp4 to video448.mp4\n",
            "Saving video449.mp4 to video449.mp4\n",
            "Saving video450.mp4 to video450.mp4\n",
            "Saving video451.mp4 to video451.mp4\n",
            "Saving video452.mp4 to video452.mp4\n",
            "Saving video453.mp4 to video453.mp4\n",
            "Saving video454.mp4 to video454.mp4\n",
            "Saving video455.mp4 to video455.mp4\n",
            "Saving video456.mp4 to video456.mp4\n",
            "Saving video457.mp4 to video457.mp4\n",
            "Saving video458.mp4 to video458.mp4\n",
            "Saving video459.mp4 to video459.mp4\n",
            "Saving video460.mp4 to video460.mp4\n",
            "Saving video461.mp4 to video461.mp4\n",
            "Saving video462.mp4 to video462.mp4\n",
            "Saving video463.mp4 to video463.mp4\n",
            "Saving video464.mp4 to video464.mp4\n",
            "Saving video465.mp4 to video465.mp4\n",
            "Saving video466.mp4 to video466.mp4\n",
            "Saving video467.mp4 to video467.mp4\n",
            "Saving video468.mp4 to video468.mp4\n",
            "Saving video469.mp4 to video469.mp4\n",
            "Saving video470.mp4 to video470.mp4\n",
            "Saving video471.mp4 to video471.mp4\n",
            "Saving video472.mp4 to video472.mp4\n",
            "Saving video473.mp4 to video473.mp4\n",
            "Saving video474.mp4 to video474.mp4\n",
            "Saving video475.mp4 to video475.mp4\n",
            "Saving video476.mp4 to video476.mp4\n",
            "Saving video477.mp4 to video477.mp4\n",
            "Saving video478.mp4 to video478.mp4\n",
            "Saving video479.mp4 to video479.mp4\n",
            "Saving video480.mp4 to video480.mp4\n",
            "Saving video481.mp4 to video481.mp4\n",
            "Saving video482.mp4 to video482.mp4\n",
            "Saving video483.mp4 to video483.mp4\n",
            "Saving video484.mp4 to video484.mp4\n",
            "Saving video485.mp4 to video485.mp4\n",
            "Saving video486.mp4 to video486.mp4\n",
            "Saving video487.mp4 to video487.mp4\n",
            "Saving video488.mp4 to video488.mp4\n",
            "Saving video489.mp4 to video489.mp4\n",
            "Saving video490.mp4 to video490.mp4\n",
            "Saving video491.mp4 to video491.mp4\n",
            "Saving video492.mp4 to video492.mp4\n",
            "Saving video493.mp4 to video493.mp4\n",
            "Saving video494.mp4 to video494.mp4\n",
            "Saving video495.mp4 to video495.mp4\n",
            "Saving video496.mp4 to video496.mp4\n",
            "Saving video497.mp4 to video497.mp4\n",
            "Saving video498.mp4 to video498.mp4\n",
            "Saving video499.mp4 to video499.mp4\n",
            "Saving video500.mp4 to video500.mp4\n",
            "Saving video501.mp4 to video501.mp4\n",
            "Saving video502.mp4 to video502.mp4\n",
            "Saving video503.mp4 to video503.mp4\n",
            "Saving video504.mp4 to video504.mp4\n",
            "Saving video505.mp4 to video505.mp4\n",
            "Saving video506.mp4 to video506.mp4\n",
            "Saving video507.mp4 to video507.mp4\n",
            "Saving video508.mp4 to video508.mp4\n",
            "Saving video509.mp4 to video509.mp4\n",
            "Saving video510.mp4 to video510.mp4\n",
            "Saving video511.mp4 to video511.mp4\n",
            "Saving video512.mp4 to video512.mp4\n",
            "Saving video513.mp4 to video513.mp4\n",
            "Saving video514.mp4 to video514.mp4\n",
            "Saving video515.mp4 to video515.mp4\n",
            "Saving video516.mp4 to video516.mp4\n",
            "Saving video517.mp4 to video517.mp4\n",
            "Saving video518.mp4 to video518.mp4\n",
            "Saving video519.mp4 to video519.mp4\n",
            "Saving video520.mp4 to video520.mp4\n",
            "Saving video521.mp4 to video521.mp4\n",
            "Saving video522.mp4 to video522.mp4\n",
            "Saving video523.mp4 to video523.mp4\n",
            "Saving video524.mp4 to video524.mp4\n",
            "Saving video525.mp4 to video525.mp4\n",
            "Saving video526.mp4 to video526.mp4\n",
            "Saving video527.mp4 to video527.mp4\n",
            "Saving video528.mp4 to video528.mp4\n",
            "Saving video529.mp4 to video529.mp4\n",
            "Saving video530.mp4 to video530.mp4\n",
            "Saving video531.mp4 to video531.mp4\n",
            "Saving video532.mp4 to video532.mp4\n",
            "Saving video533.mp4 to video533.mp4\n",
            "Saving video534.mp4 to video534.mp4\n",
            "Saving video535.mp4 to video535.mp4\n",
            "Saving video536.mp4 to video536.mp4\n",
            "Saving video537.mp4 to video537.mp4\n",
            "Saving video538.mp4 to video538.mp4\n",
            "Saving video539.mp4 to video539.mp4\n",
            "Saving video540.mp4 to video540.mp4\n",
            "Saving video541.mp4 to video541.mp4\n",
            "Saving video542.mp4 to video542.mp4\n",
            "Saving video543.mp4 to video543.mp4\n",
            "Saving video544.mp4 to video544.mp4\n",
            "Saving video545.mp4 to video545.mp4\n",
            "Saving video546.mp4 to video546.mp4\n",
            "Saving video547.mp4 to video547.mp4\n",
            "Saving video548.mp4 to video548.mp4\n",
            "Saving video549.mp4 to video549.mp4\n",
            "Saving video550.mp4 to video550.mp4\n",
            "Saving video551.mp4 to video551.mp4\n",
            "Saving video552.mp4 to video552.mp4\n",
            "Saving video553.mp4 to video553.mp4\n",
            "Saving video554.mp4 to video554.mp4\n",
            "Saving video555.mp4 to video555.mp4\n",
            "Saving video556.mp4 to video556.mp4\n",
            "Saving video557.mp4 to video557.mp4\n",
            "Saving video558.mp4 to video558.mp4\n",
            "Saving video559.mp4 to video559.mp4\n",
            "Saving video560.mp4 to video560.mp4\n",
            "Saving video561.mp4 to video561.mp4\n",
            "Saving video562.mp4 to video562.mp4\n",
            "Saving video563.mp4 to video563.mp4\n",
            "Saving video564.mp4 to video564.mp4\n",
            "Saving video565.mp4 to video565.mp4\n",
            "Saving video566.mp4 to video566.mp4\n",
            "Saving video567.mp4 to video567.mp4\n",
            "Saving video568.mp4 to video568.mp4\n",
            "Saving video569.mp4 to video569.mp4\n",
            "Saving video570.mp4 to video570.mp4\n",
            "Saving video571.mp4 to video571.mp4\n",
            "Saving video572.mp4 to video572.mp4\n",
            "Saving video573.mp4 to video573.mp4\n",
            "Saving video574.mp4 to video574.mp4\n",
            "Saving video575.mp4 to video575.mp4\n",
            "Saving video576.mp4 to video576.mp4\n",
            "Saving video577.mp4 to video577.mp4\n",
            "Saving video578.mp4 to video578.mp4\n",
            "Saving video579.mp4 to video579.mp4\n",
            "Saving video580.mp4 to video580.mp4\n",
            "Saving video581.mp4 to video581.mp4\n",
            "Saving video582.mp4 to video582.mp4\n",
            "Saving video583.mp4 to video583.mp4\n",
            "Saving video584.mp4 to video584.mp4\n",
            "Saving video585.mp4 to video585.mp4\n",
            "Saving video586.mp4 to video586.mp4\n",
            "Saving video587.mp4 to video587.mp4\n",
            "Saving video588.mp4 to video588.mp4\n",
            "Saving video589.mp4 to video589.mp4\n",
            "Saving video590.mp4 to video590.mp4\n",
            "Saving video591.mp4 to video591.mp4\n",
            "Saving video592.mp4 to video592.mp4\n",
            "Saving video593.mp4 to video593.mp4\n",
            "Saving video594.mp4 to video594.mp4\n",
            "Saving video595.mp4 to video595.mp4\n",
            "Saving video596.mp4 to video596.mp4\n",
            "Saving video597.mp4 to video597.mp4\n",
            "Saving video598.mp4 to video598.mp4\n",
            "Saving video599.mp4 to video599.mp4\n",
            "Saving video600.mp4 to video600.mp4\n",
            "Saving video601.mp4 to video601.mp4\n",
            "Saving video602.mp4 to video602.mp4\n",
            "Saving video603.mp4 to video603.mp4\n",
            "Saving video604.mp4 to video604.mp4\n",
            "Saving video605.mp4 to video605.mp4\n",
            "Saving video606.mp4 to video606.mp4\n",
            "Saving video607.mp4 to video607.mp4\n",
            "Saving video608.mp4 to video608.mp4\n",
            "Saving video609.mp4 to video609.mp4\n",
            "Saving video610.mp4 to video610.mp4\n",
            "Saving video611.mp4 to video611.mp4\n",
            "Saving video612.mp4 to video612.mp4\n",
            "Saving video613.mp4 to video613.mp4\n",
            "Saving video614.mp4 to video614.mp4\n",
            "Saving video615.mp4 to video615.mp4\n",
            "Saving video616.mp4 to video616.mp4\n",
            "Saving video617.mp4 to video617.mp4\n",
            "Saving video618.mp4 to video618.mp4\n",
            "Saving video619.mp4 to video619.mp4\n",
            "Saving video620.mp4 to video620.mp4\n",
            "Saving video621.mp4 to video621.mp4\n",
            "Saving video622.mp4 to video622.mp4\n",
            "Saving video623.mp4 to video623.mp4\n",
            "Saving video624.mp4 to video624.mp4\n",
            "Saving video625.mp4 to video625.mp4\n",
            "Saving video626.mp4 to video626.mp4\n",
            "Saving video627.mp4 to video627.mp4\n",
            "Saving video628.mp4 to video628.mp4\n",
            "Saving video629.mp4 to video629.mp4\n",
            "Saving video630.mp4 to video630.mp4\n",
            "Saving video631.mp4 to video631.mp4\n",
            "Saving video632.mp4 to video632.mp4\n",
            "Saving video633.mp4 to video633.mp4\n",
            "Saving video634.mp4 to video634.mp4\n",
            "Saving video635.mp4 to video635.mp4\n",
            "Saving video636.mp4 to video636.mp4\n",
            "Saving video637.mp4 to video637.mp4\n",
            "Saving video638.mp4 to video638.mp4\n",
            "Saving video639.mp4 to video639.mp4\n",
            "Saving video640.mp4 to video640.mp4\n",
            "Saving video641.mp4 to video641.mp4\n",
            "Saving video642.mp4 to video642.mp4\n",
            "Saving video643.mp4 to video643.mp4\n",
            "Saving video644.mp4 to video644.mp4\n",
            "Saving video645.mp4 to video645.mp4\n",
            "Saving video646.mp4 to video646.mp4\n",
            "Saving video647.mp4 to video647.mp4\n",
            "Saving video648.mp4 to video648.mp4\n",
            "Saving video649.mp4 to video649.mp4\n",
            "Saving video650.mp4 to video650.mp4\n",
            "Saving video651.mp4 to video651.mp4\n",
            "Saving video652.mp4 to video652.mp4\n",
            "Saving video653.mp4 to video653.mp4\n",
            "Saving video654.mp4 to video654.mp4\n",
            "Saving video655.mp4 to video655.mp4\n",
            "Saving video656.mp4 to video656.mp4\n",
            "Saving video657.mp4 to video657.mp4\n",
            "Saving video658.mp4 to video658.mp4\n",
            "Saving video659.mp4 to video659.mp4\n",
            "Saving video660.mp4 to video660.mp4\n",
            "Saving video661.mp4 to video661.mp4\n",
            "Saving video662.mp4 to video662.mp4\n",
            "Saving video663.mp4 to video663.mp4\n",
            "Saving video664.mp4 to video664.mp4\n",
            "Saving video665.mp4 to video665.mp4\n",
            "Saving video666.mp4 to video666.mp4\n",
            "Saving video667.mp4 to video667.mp4\n",
            "Saving video668.mp4 to video668.mp4\n",
            "Saving video669.mp4 to video669.mp4\n",
            "Saving video670.mp4 to video670.mp4\n",
            "Saving video671.mp4 to video671.mp4\n",
            "Saving video672.mp4 to video672.mp4\n",
            "Saving video673.mp4 to video673.mp4\n",
            "Saving video674.mp4 to video674.mp4\n",
            "Saving video675.mp4 to video675.mp4\n",
            "Saving video676.mp4 to video676.mp4\n",
            "Saving video677.mp4 to video677.mp4\n",
            "Saving video678.mp4 to video678.mp4\n",
            "Saving video679.mp4 to video679.mp4\n",
            "Saving video680.mp4 to video680.mp4\n",
            "Saving video681.mp4 to video681.mp4\n",
            "Saving video682.mp4 to video682.mp4\n",
            "Saving video683.mp4 to video683.mp4\n",
            "Saving video684.mp4 to video684.mp4\n",
            "Saving video685.mp4 to video685.mp4\n",
            "Saving video686.mp4 to video686.mp4\n",
            "Saving video687.mp4 to video687.mp4\n",
            "Saving video688.mp4 to video688.mp4\n",
            "Saving video689.mp4 to video689.mp4\n",
            "Saving video690.mp4 to video690.mp4\n",
            "Saving video691.mp4 to video691.mp4\n",
            "Saving video692.mp4 to video692.mp4\n",
            "Saving video693.mp4 to video693.mp4\n",
            "Saving video694.mp4 to video694.mp4\n",
            "Saving video695.mp4 to video695.mp4\n",
            "Saving video696.mp4 to video696.mp4\n",
            "Saving video697.mp4 to video697.mp4\n",
            "Saving video698.mp4 to video698.mp4\n",
            "Saving video699.mp4 to video699.mp4\n",
            "Saving video700.mp4 to video700.mp4\n",
            "Saving video701.mp4 to video701.mp4\n",
            "Saving video702.mp4 to video702.mp4\n",
            "Saving video703.mp4 to video703.mp4\n",
            "Saving video704.mp4 to video704.mp4\n",
            "Saving video705.mp4 to video705.mp4\n",
            "Saving video706.mp4 to video706.mp4\n",
            "Saving video707.mp4 to video707.mp4\n",
            "Saving video708.mp4 to video708.mp4\n",
            "Saving video709.mp4 to video709.mp4\n",
            "Saving video710.mp4 to video710.mp4\n",
            "Saving video711.mp4 to video711.mp4\n",
            "Saving video712.mp4 to video712.mp4\n",
            "Saving video713.mp4 to video713.mp4\n",
            "Saving video714.mp4 to video714.mp4\n",
            "Saving video715.mp4 to video715.mp4\n",
            "Saving video716.mp4 to video716.mp4\n",
            "Saving video717.mp4 to video717.mp4\n",
            "Saving video718.mp4 to video718.mp4\n",
            "Saving video719.mp4 to video719.mp4\n",
            "Saving video720.mp4 to video720.mp4\n",
            "Saving video721.mp4 to video721.mp4\n",
            "Saving video722.mp4 to video722.mp4\n",
            "Saving video723.mp4 to video723.mp4\n",
            "Saving video724.mp4 to video724.mp4\n",
            "Saving video725.mp4 to video725.mp4\n",
            "Saving video726.mp4 to video726.mp4\n",
            "Saving video727.mp4 to video727.mp4\n",
            "Saving video728.mp4 to video728.mp4\n",
            "Saving video729.mp4 to video729.mp4\n",
            "Saving video730.mp4 to video730.mp4\n",
            "Saving video731.mp4 to video731.mp4\n",
            "Saving video732.mp4 to video732.mp4\n",
            "Saving video733.mp4 to video733.mp4\n",
            "Saving video734.mp4 to video734.mp4\n",
            "Saving video735.mp4 to video735.mp4\n",
            "Saving video736.mp4 to video736.mp4\n",
            "Saving video737.mp4 to video737.mp4\n",
            "Saving video738.mp4 to video738.mp4\n",
            "Saving video739.mp4 to video739.mp4\n",
            "Saving video740.mp4 to video740.mp4\n",
            "Saving video741.mp4 to video741.mp4\n",
            "Saving video742.mp4 to video742.mp4\n",
            "Saving video743.mp4 to video743.mp4\n",
            "Saving video744.mp4 to video744.mp4\n",
            "Saving video745.mp4 to video745.mp4\n",
            "Saving video746.mp4 to video746.mp4\n",
            "Saving video747.mp4 to video747.mp4\n",
            "Saving video748.mp4 to video748.mp4\n",
            "Saving video749.mp4 to video749.mp4\n",
            "Saving video750.mp4 to video750.mp4\n",
            "Saving video751.mp4 to video751.mp4\n",
            "Saving video752.mp4 to video752.mp4\n",
            "Saving video753.mp4 to video753.mp4\n",
            "Saving video754.mp4 to video754.mp4\n",
            "Saving video755.mp4 to video755.mp4\n",
            "Saving video756.mp4 to video756.mp4\n",
            "Saving video757.mp4 to video757.mp4\n",
            "Saving video758.mp4 to video758.mp4\n",
            "Saving video759.mp4 to video759.mp4\n",
            "Saving video760.mp4 to video760.mp4\n",
            "Saving video761.mp4 to video761.mp4\n",
            "Saving video762.mp4 to video762.mp4\n",
            "Saving video763.mp4 to video763.mp4\n",
            "Saving video764.mp4 to video764.mp4\n",
            "Saving video765.mp4 to video765.mp4\n",
            "Saving video766.mp4 to video766.mp4\n",
            "Saving video767.mp4 to video767.mp4\n",
            "Saving video768.mp4 to video768.mp4\n",
            "Saving video769.mp4 to video769.mp4\n",
            "Saving video770.mp4 to video770.mp4\n",
            "Saving video771.mp4 to video771.mp4\n",
            "Saving video772.mp4 to video772.mp4\n",
            "Saving video773.mp4 to video773.mp4\n",
            "Saving video774.mp4 to video774.mp4\n",
            "Saving video775.mp4 to video775.mp4\n",
            "Saving video776.mp4 to video776.mp4\n",
            "Saving video777.mp4 to video777.mp4\n",
            "Saving video778.mp4 to video778.mp4\n",
            "Saving video779.mp4 to video779.mp4\n",
            "Saving video780.mp4 to video780.mp4\n",
            "Saving video781.mp4 to video781.mp4\n",
            "Saving video782.mp4 to video782.mp4\n",
            "Saving video783.mp4 to video783.mp4\n",
            "Saving video784.mp4 to video784.mp4\n",
            "Saving video785.mp4 to video785.mp4\n",
            "Saving video786.mp4 to video786.mp4\n",
            "Saving video787.mp4 to video787.mp4\n",
            "Saving video788.mp4 to video788.mp4\n",
            "Saving video789.mp4 to video789.mp4\n",
            "Saving video790.mp4 to video790.mp4\n",
            "Saving video791.mp4 to video791.mp4\n",
            "Saving video792.mp4 to video792.mp4\n",
            "Saving video793.mp4 to video793.mp4\n",
            "Saving video794.mp4 to video794.mp4\n",
            "Saving video795.mp4 to video795.mp4\n",
            "Saving video796.mp4 to video796.mp4\n",
            "Saving video797.mp4 to video797.mp4\n",
            "Saving video798.mp4 to video798.mp4\n",
            "Saving video799.mp4 to video799.mp4\n",
            "Saving video800.mp4 to video800.mp4\n",
            "Saving video801.mp4 to video801.mp4\n",
            "Saving video802.mp4 to video802.mp4\n",
            "Saving video803.mp4 to video803.mp4\n",
            "Saving video804.mp4 to video804.mp4\n",
            "Saving video805.mp4 to video805.mp4\n",
            "Saving video806.mp4 to video806.mp4\n",
            "Saving video807.mp4 to video807.mp4\n",
            "Saving video808.mp4 to video808.mp4\n",
            "Saving video809.mp4 to video809.mp4\n",
            "Saving video810.mp4 to video810.mp4\n",
            "Saving video811.mp4 to video811.mp4\n",
            "Saving video812.mp4 to video812.mp4\n",
            "Saving video813.mp4 to video813.mp4\n",
            "Saving video814.mp4 to video814.mp4\n",
            "Saving video815.mp4 to video815.mp4\n",
            "Saving video816.mp4 to video816.mp4\n",
            "Saving video817.mp4 to video817.mp4\n",
            "Saving video818.mp4 to video818.mp4\n",
            "Saving video819.mp4 to video819.mp4\n",
            "Saving video820.mp4 to video820.mp4\n",
            "Saving video821.mp4 to video821.mp4\n",
            "Saving video822.mp4 to video822.mp4\n",
            "Saving video823.mp4 to video823.mp4\n",
            "Saving video824.mp4 to video824.mp4\n",
            "Saving video825.mp4 to video825.mp4\n",
            "Saving video826.mp4 to video826.mp4\n",
            "Saving video827.mp4 to video827.mp4\n",
            "Saving video828.mp4 to video828.mp4\n",
            "Saving video829.mp4 to video829.mp4\n",
            "Saving video830.mp4 to video830.mp4\n",
            "Saving video831.mp4 to video831.mp4\n",
            "Saving video832.mp4 to video832.mp4\n",
            "Saving video833.mp4 to video833.mp4\n",
            "Saving video834.mp4 to video834.mp4\n",
            "Saving video835.mp4 to video835.mp4\n",
            "Saving video836.mp4 to video836.mp4\n",
            "Saving video837.mp4 to video837.mp4\n",
            "Saving video838.mp4 to video838.mp4\n",
            "Saving video839.mp4 to video839.mp4\n",
            "Saving video840.mp4 to video840.mp4\n",
            "Saving video841.mp4 to video841.mp4\n",
            "Saving video842.mp4 to video842.mp4\n",
            "Saving video843.mp4 to video843.mp4\n",
            "Saving video844.mp4 to video844.mp4\n",
            "Saving video845.mp4 to video845.mp4\n",
            "Saving video846.mp4 to video846.mp4\n",
            "Saving video847.mp4 to video847.mp4\n",
            "Saving video848.mp4 to video848.mp4\n",
            "Saving video849.mp4 to video849.mp4\n",
            "Saving video850.mp4 to video850.mp4\n",
            "Saving video851.mp4 to video851.mp4\n",
            "Saving video852.mp4 to video852.mp4\n",
            "Saving video853.mp4 to video853.mp4\n",
            "Saving video854.mp4 to video854.mp4\n",
            "Saving video855.mp4 to video855.mp4\n",
            "Saving video856.mp4 to video856.mp4\n",
            "Saving video857.mp4 to video857.mp4\n",
            "Saving video858.mp4 to video858.mp4\n",
            "Saving video859.mp4 to video859.mp4\n",
            "Saving video860.mp4 to video860.mp4\n",
            "Saving video861.mp4 to video861.mp4\n",
            "Saving video862.mp4 to video862.mp4\n",
            "Saving video863.mp4 to video863.mp4\n",
            "Saving video864.mp4 to video864.mp4\n",
            "Saving video865.mp4 to video865.mp4\n",
            "Saving video866.mp4 to video866.mp4\n",
            "Saving video867.mp4 to video867.mp4\n",
            "Saving video868.mp4 to video868.mp4\n",
            "Saving video869.mp4 to video869.mp4\n",
            "Saving video870.mp4 to video870.mp4\n",
            "Saving video871.mp4 to video871.mp4\n",
            "Saving video872.mp4 to video872.mp4\n",
            "Saving video873.mp4 to video873.mp4\n",
            "Saving video874.mp4 to video874.mp4\n",
            "Saving video875.mp4 to video875.mp4\n",
            "Saving video876.mp4 to video876.mp4\n",
            "Saving video877.mp4 to video877.mp4\n",
            "Saving video878.mp4 to video878.mp4\n",
            "Saving video879.mp4 to video879.mp4\n",
            "Saving video880.mp4 to video880.mp4\n",
            "Saving video881.mp4 to video881.mp4\n",
            "Saving video882.mp4 to video882.mp4\n",
            "Saving video883.mp4 to video883.mp4\n",
            "Saving video884.mp4 to video884.mp4\n",
            "Saving video885.mp4 to video885.mp4\n",
            "Saving video886.mp4 to video886.mp4\n",
            "Saving video887.mp4 to video887.mp4\n",
            "Saving video888.mp4 to video888.mp4\n",
            "Saving video889.mp4 to video889.mp4\n",
            "Saving video890.mp4 to video890.mp4\n",
            "Saving video891.mp4 to video891.mp4\n",
            "Saving video892.mp4 to video892.mp4\n",
            "Saving video893.mp4 to video893.mp4\n",
            "Saving video894.mp4 to video894.mp4\n",
            "Saving video895.mp4 to video895.mp4\n",
            "Saving video896.mp4 to video896.mp4\n",
            "Saving video897.mp4 to video897.mp4\n",
            "Saving video898.mp4 to video898.mp4\n",
            "Saving video899.mp4 to video899.mp4\n",
            "Saving video900.mp4 to video900.mp4\n",
            "Saving video901.mp4 to video901.mp4\n",
            "Saving video902.mp4 to video902.mp4\n",
            "Saving video903.mp4 to video903.mp4\n",
            "Saving video904.mp4 to video904.mp4\n",
            "Saving video905.mp4 to video905.mp4\n",
            "Saving video906.mp4 to video906.mp4\n",
            "Saving video907.mp4 to video907.mp4\n",
            "Saving video908.mp4 to video908.mp4\n",
            "Saving video909.mp4 to video909.mp4\n",
            "Saving video910.mp4 to video910.mp4\n",
            "Saving video911.mp4 to video911.mp4\n",
            "Saving video912.mp4 to video912.mp4\n",
            "Saving video913.mp4 to video913.mp4\n",
            "Saving video914.mp4 to video914.mp4\n",
            "Saving video915.mp4 to video915.mp4\n",
            "Saving video916.mp4 to video916.mp4\n",
            "Saving video917.mp4 to video917.mp4\n",
            "Saving video918.mp4 to video918.mp4\n",
            "Saving video919.mp4 to video919.mp4\n",
            "Saving video920.mp4 to video920.mp4\n",
            "Saving video921.mp4 to video921.mp4\n",
            "Saving video922.mp4 to video922.mp4\n",
            "Saving video923.mp4 to video923.mp4\n",
            "Saving video924.mp4 to video924.mp4\n",
            "Saving video925.mp4 to video925.mp4\n",
            "Saving video926.mp4 to video926.mp4\n",
            "Saving video927.mp4 to video927.mp4\n",
            "Saving video928.mp4 to video928.mp4\n",
            "Saving video929.mp4 to video929.mp4\n",
            "Saving video930.mp4 to video930.mp4\n",
            "Saving video931.mp4 to video931.mp4\n",
            "Saving video932.mp4 to video932.mp4\n",
            "Saving video933.mp4 to video933.mp4\n",
            "Saving video934.mp4 to video934.mp4\n",
            "Saving video935.mp4 to video935.mp4\n",
            "Saving video936.mp4 to video936.mp4\n",
            "Saving video937.mp4 to video937.mp4\n",
            "Saving video938.mp4 to video938.mp4\n",
            "Saving video939.mp4 to video939.mp4\n",
            "Saving video940.mp4 to video940.mp4\n",
            "Saving video941.mp4 to video941.mp4\n",
            "Saving video942.mp4 to video942.mp4\n",
            "Saving video943.mp4 to video943.mp4\n",
            "Saving video944.mp4 to video944.mp4\n",
            "Saving video945.mp4 to video945.mp4\n",
            "Saving video946.mp4 to video946.mp4\n",
            "Saving video947.mp4 to video947.mp4\n",
            "Saving video948.mp4 to video948.mp4\n",
            "Saving video949.mp4 to video949.mp4\n",
            "Saving video950.mp4 to video950.mp4\n",
            "Saving video951.mp4 to video951.mp4\n",
            "Saving video952.mp4 to video952.mp4\n",
            "Saving video953.mp4 to video953.mp4\n",
            "Saving video954.mp4 to video954.mp4\n",
            "Saving video955.mp4 to video955.mp4\n",
            "Saving video956.mp4 to video956.mp4\n",
            "Saving video957.mp4 to video957.mp4\n",
            "Saving video958.mp4 to video958.mp4\n",
            "Saving video959.mp4 to video959.mp4\n",
            "Saving video960.mp4 to video960.mp4\n",
            "Saving video961.mp4 to video961.mp4\n",
            "Saving video962.mp4 to video962.mp4\n",
            "Saving video963.mp4 to video963.mp4\n",
            "Saving video964.mp4 to video964.mp4\n",
            "Saving video965.mp4 to video965.mp4\n",
            "Saving video966.mp4 to video966.mp4\n",
            "Saving video967.mp4 to video967.mp4\n",
            "Saving video968.mp4 to video968.mp4\n",
            "Saving video969.mp4 to video969.mp4\n",
            "Saving video970.mp4 to video970.mp4\n",
            "Saving video971.mp4 to video971.mp4\n",
            "Saving video972.mp4 to video972.mp4\n",
            "Saving video973.mp4 to video973.mp4\n",
            "Saving video974.mp4 to video974.mp4\n",
            "Saving video975.mp4 to video975.mp4\n",
            "Saving video976.mp4 to video976.mp4\n",
            "Saving video977.mp4 to video977.mp4\n",
            "Saving video978.mp4 to video978.mp4\n",
            "Saving video979.mp4 to video979.mp4\n",
            "Saving video980.mp4 to video980.mp4\n",
            "Saving video981.mp4 to video981.mp4\n",
            "Saving video982.mp4 to video982.mp4\n",
            "Saving video983.mp4 to video983.mp4\n",
            "Saving video984.mp4 to video984.mp4\n",
            "Saving video985.mp4 to video985.mp4\n",
            "Saving video986.mp4 to video986.mp4\n",
            "Saving video987.mp4 to video987.mp4\n",
            "Saving video988.mp4 to video988.mp4\n",
            "Saving video989.mp4 to video989.mp4\n",
            "Saving video990.mp4 to video990.mp4\n",
            "Saving video991.mp4 to video991.mp4\n",
            "Saving video992.mp4 to video992.mp4\n",
            "Saving video993.mp4 to video993.mp4\n",
            "Saving video996.mp4 to video996.mp4\n",
            "Saving video997.mp4 to video997.mp4\n",
            "Saving video998.mp4 to video998.mp4\n",
            "Saving video999.mp4 to video999.mp4\n",
            "Saving video1000.mp4 to video1000.mp4\n",
            "Saving video1001.mp4 to video1001.mp4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "pip install opencv-python"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gM-vP6jd6MeQ",
        "outputId": "f63feee3-13a2-4e9c-d2ce-f7e975648773"
      },
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: opencv-python in /usr/local/lib/python3.10/dist-packages (4.10.0.84)\n",
            "Requirement already satisfied: numpy>=1.21.2 in /usr/local/lib/python3.10/dist-packages (from opencv-python) (1.26.4)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import cv2\n",
        "import os\n",
        "\n",
        "# Specify the path to your video file\n",
        "video_path = 'videos/video1.mp4'  # Adjust this path accordingly\n",
        "\n",
        "# Print absolute path for debugging\n",
        "print(f\"Attempting to open video file at: {os.path.abspath(video_path)}\")\n",
        "\n",
        "# Create a VideoCapture object\n",
        "cap = cv2.VideoCapture(video_path)\n",
        "\n",
        "# Check if the video opened successfully\n",
        "if not cap.isOpened():\n",
        "    print(f\"Could not open video: {video_path}\")\n",
        "else:\n",
        "    print(\"Video opened successfully.\")\n",
        "\n",
        "# Read and display the video frame by frame\n",
        "while cap.isOpened():\n",
        "    ret, frame = cap.read()\n",
        "    if not ret:\n",
        "        print(\"End of video or cannot read the video frame.\")\n",
        "        break\n",
        "\n",
        "    # Display the frame in a window\n",
        "    cv2.imshow('Video', frame)\n",
        "\n",
        "    # Press 'q' on the keyboard to exit the video early\n",
        "    if cv2.waitKey(25) & 0xFF == ord('q'):\n",
        "        break\n",
        "\n",
        "# Release the video capture object and close all OpenCV windows\n",
        "cap.release()\n",
        "cv2.destroyAllWindows()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QWHAMddv6XaA",
        "outputId": "67df0d93-e65b-4c67-8255-6f2a4c7137e4"
      },
      "execution_count": 32,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Attempting to open video file at: /content/videos/video1.mp4\n",
            "Could not open video: videos/video1.mp4\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import os\n",
        "import gradio as gr\n",
        "import cv2\n",
        "from PIL import Image\n",
        "import numpy as np\n",
        "\n",
        "# Sample video database\n",
        "video_database = [\n",
        "    {\"path\": \"MSRVTT/video1.mp4\", \"tags\": [\"cooking\", \"food\"]},\n",
        "    {\"path\": \"MSRVTT/video2.mp4\", \"tags\": [\"dancing\", \"party\", \"performance\"]},\n",
        "    {\"path\": \"MSRVTT/video3.mp4\", \"tags\": [\"sports\", \"exercise\", \"game\"]},\n",
        "]"
      ],
      "metadata": {
        "id": "JU7nZ6tQb88I"
      },
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def process_video_for_display(video_path, max_frames=6):\n",
        "    try:\n",
        "        cap = cv2.VideoCapture(video_path)\n",
        "        if not cap.isOpened():\n",
        "            print(f\"Error: Could not open video '{video_path}'.\")\n",
        "            return []\n",
        "\n",
        "        total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))\n",
        "        if total_frames == 0:\n",
        "            print(f\"Video '{video_path}' has no frames.\")\n",
        "            return []\n",
        "\n",
        "        # Ensure we have at least 1 frame\n",
        "        if total_frames < max_frames:\n",
        "            max_frames = total_frames\n",
        "\n",
        "        indices = np.linspace(0, total_frames-1, max_frames, dtype=int)\n",
        "\n",
        "        frames = []\n",
        "        for idx in indices:\n",
        "            cap.set(cv2.CAP_PROP_POS_FRAMES, idx)\n",
        "            ret, frame = cap.read()\n",
        "            if ret:\n",
        "                frame = cv2.cvtColor(frame, cv2.COLOR_BGR2RGB)\n",
        "                frames.append(Image.fromarray(frame))\n",
        "            else:\n",
        "                print(f\"Error reading frame {idx} from video '{video_path}'.\")\n",
        "\n",
        "        cap.release()\n",
        "        return frames\n",
        "    except Exception as e:\n",
        "        print(f\"Error processing video '{video_path}': {str(e)}\")\n",
        "        return []"
      ],
      "metadata": {
        "id": "k_lHGzVJ8ZNW"
      },
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def retrieve_video_from_larger_set(query, num_videos=1, top_k=5):\n",
        "    relevant_videos = []\n",
        "    query_tags = query.lower().split()  # Tokenize the input query\n",
        "\n",
        "    print(f\"Query: {query} | Query Tags: {query_tags}\")  # Debugging purpose\n",
        "\n",
        "    for video in video_database:\n",
        "        # Check if any of the query tags are in the video's tags\n",
        "        if any(tag in video['tags'] for tag in query_tags):\n",
        "            relevant_videos.append((video['path'], 1.0))  # Mock similarity score\n",
        "            print(f\"Match found: {video}\")  # Debugging\n",
        "\n",
        "    # Sort by similarity score (if there were real scores) and take the top k results\n",
        "    relevant_videos.sort(key=lambda x: x[1], reverse=True)\n",
        "    return relevant_videos[:top_k]"
      ],
      "metadata": {
        "id": "5pFx9QfL8byt"
      },
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "def video_search(query, num_results=5):\n",
        "    try:\n",
        "        results = retrieve_video_from_larger_set(query, top_k=num_results)\n",
        "\n",
        "        if not results:\n",
        "            return [], \"No results found.\"\n",
        "\n",
        "        output_frames = []\n",
        "        captions = []\n",
        "\n",
        "        for video_path, similarity in results:\n",
        "            frames = process_video_for_display(video_path)\n",
        "            if frames:\n",
        "                output_frames.extend(frames)\n",
        "                video_name = os.path.basename(video_path)\n",
        "                captions.extend([f\"{video_name} (similarity: {similarity:.3f})\"] * len(frames))\n",
        "            else:\n",
        "                print(f\"Warning: No frames extracted from video '{video_path}'.\")\n",
        "\n",
        "        if not output_frames:\n",
        "            return [], \"No frames could be extracted.\"\n",
        "\n",
        "        return output_frames, f\"Found {len(results)} matching videos.\"\n",
        "\n",
        "    except Exception as e:\n",
        "        return [], f\"Error: {str(e)}\""
      ],
      "metadata": {
        "id": "byMGRWin8eg1"
      },
      "execution_count": 44,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Example usage\n",
        "output_frames, status = video_search(\"cooking\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bsPtXVah8_vb",
        "outputId": "af044469-4908-4ad9-d28d-c9480a8c3f5a"
      },
      "execution_count": 45,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Query: cooking | Query Tags: ['cooking']\n",
            "Match found: {'path': 'videos/video1.mp4', 'tags': ['cooking', 'food']}\n",
            "Error: Could not open video 'videos/video1.mp4'.\n",
            "Warning: No frames extracted from video 'videos/video1.mp4'.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Gradio Interface\n",
        "with gr.Blocks(title=\"Video Search\", theme=gr.themes.Base()) as demo:\n",
        "    gr.Markdown(\"\"\"\n",
        "    # Video Search using CRET Model\n",
        "    Search through videos using natural language descriptions.\n",
        "    \"\"\")\n",
        "\n",
        "    with gr.Row():\n",
        "        with gr.Column(scale=3):\n",
        "            query_input = gr.Textbox(\n",
        "                label=\"Enter your text query\",\n",
        "                placeholder=\"Example: a person is cooking food\",\n",
        "                lines=2\n",
        "            )\n",
        "\n",
        "            example_queries = gr.Examples(\n",
        "                examples=[\n",
        "                    \"cooking\",\n",
        "                    \"dancing\",\n",
        "                    \"sports\"\n",
        "                ],\n",
        "                inputs=query_input,\n",
        "                label=\"Example Queries\"\n",
        "            )\n",
        "\n",
        "        with gr.Column(scale=1):\n",
        "            num_results = gr.Slider(\n",
        "                minimum=1,\n",
        "                maximum=10,\n",
        "                value=5,\n",
        "                step=1,\n",
        "                label=\"Number of results\"\n",
        "            )\n",
        "\n",
        "    search_button = gr.Button(\"Search\", variant=\"primary\")\n",
        "\n",
        "    status = gr.Textbox(label=\"Status\", value=\"Ready\", interactive=False)\n",
        "\n",
        "    gallery = gr.Gallery(\n",
        "        label=\"Results\",\n",
        "        show_label=True,\n",
        "        columns=3,\n",
        "        rows=None,\n",
        "        height=\"500px\"\n",
        "    )\n",
        "\n",
        "    search_button.click(\n",
        "        fn=video_search,\n",
        "        inputs=[query_input, num_results],\n",
        "        outputs=[gallery, status]\n",
        "    )\n",
        "\n",
        "# Launch the app\n",
        "demo.launch(debug=True, share=True)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 698
        },
        "id": "T6hMut338hE9",
        "outputId": "211eef66-7b62-4dc3-ab4f-754d5a87fb5c"
      },
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Colab notebook detected. This cell will run indefinitely so that you can see errors and logs. To turn off, set debug=False in launch().\n",
            "* Running on public URL: https://1eb81cb5e4cb600e62.gradio.live\n",
            "\n",
            "This share link expires in 72 hours. For free permanent hosting and GPU upgrades, run `gradio deploy` from the terminal in the working directory to deploy to Hugging Face Spaces (https://huggingface.co/spaces)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ],
            "text/html": [
              "<div><iframe src=\"https://1eb81cb5e4cb600e62.gradio.live\" width=\"100%\" height=\"500\" allow=\"autoplay; camera; microphone; clipboard-read; clipboard-write;\" frameborder=\"0\" allowfullscreen></iframe></div>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Query: car | Query Tags: ['car']\n",
            "Query: dancing | Query Tags: ['dancing']\n",
            "Match found: {'path': 'MSRVTT/video2.mp4', 'tags': ['dancing', 'party', 'performance']}\n",
            "Keyboard interruption in main thread... closing server.\n",
            "Killing tunnel 127.0.0.1:7861 <> https://1eb81cb5e4cb600e62.gradio.live\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": []
          },
          "metadata": {},
          "execution_count": 48
        }
      ]
    }
  ],
  "metadata": {
    "colab": {
      "provenance": []
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}