Release v3.1.0 (#10572)

Author: hhaAndroid

Diff for: .circleci/test.yml

@@ -62,6 +62,7 @@ jobs:
             equal: ["3.9.0", << parameters.python >>]
           steps:
             - run: pip install "protobuf <= 3.20.1" && sudo apt-get update && sudo apt-get -y install libprotobuf-dev protobuf-compiler cmake
+            - run: pip install dsdl
       - run:
           name: Install mmdet dependencies
           # numpy may be downgraded after building pycocotools, which causes `ImportError: numpy.core.multiarray failed to import`
@@ -73,7 +74,9 @@ jobs:
             pip install -r requirements/tests.txt -r requirements/optional.txt
             pip install --force-reinstall pycocotools
             pip install albumentations>=0.3.2 --no-binary imgaug,albumentations
+            pip install -r requirements/tracking.txt
             pip install git+https://github.com/cocodataset/panopticapi.git
+            pip install git+https://github.com/JonathonLuiten/TrackEval.git
       - run:
           name: Build and install
           command: |
@@ -91,10 +94,10 @@ jobs:
         type: string
       cuda:
         type: enum
-        enum: ["10.1", "10.2", "11.1", "11.7"]
+        enum: ["11.1", "11.7"]
       cudnn:
         type: integer
-        default: 7
+        default: 8
     machine:
       image: ubuntu-2004-cuda-11.4:202110-01
       # docker_layer_caching: true
@@ -121,7 +124,9 @@ jobs:
             docker exec mmdetection pip install -r requirements/tests.txt -r requirements/optional.txt
             docker exec mmdetection pip install pycocotools
             docker exec mmdetection pip install albumentations>=0.3.2 --no-binary imgaug,albumentations
+            docker exec mmdetection pip install -r requirements/tracking.txt
             docker exec mmdetection pip install git+https://github.com/cocodataset/panopticapi.git
+            docker exec mmdetection pip install git+https://github.com/JonathonLuiten/TrackEval.git
             docker exec mmdetection python -c 'import mmcv; print(mmcv.__version__)'
       - run:
           name: Build and install
@@ -154,9 +159,9 @@ workflows:
                 - dev-3.x
       - build_cpu:
           name: minimum_version_cpu
-          torch: 1.6.0
-          torchvision: 0.7.0
-          python: 3.7.4 # The lowest python 3.7.x version available on CircleCI images
+          torch: 1.8.0
+          torchvision: 0.9.0
+          python: 3.7.16
           requires:
             - lint
       - build_cpu:
@@ -175,7 +180,7 @@ workflows:
           torch: 1.8.1
           # Use double quotation mark to explicitly specify its type
           # as string instead of number
-          cuda: "10.2"
+          cuda: "11.1"
           requires:
             - hold
       - build_cuda:
@@ -191,8 +196,8 @@ workflows:
     jobs:
       - build_cuda:
           name: minimum_version_gpu
-          torch: 1.6.0
-          cuda: "10.1"
+          torch: 1.8.0
+          cuda: "11.1"
           filters:
             branches:
               only:

Diff for: .dev_scripts/benchmark_full_models.txt

@@ -33,12 +33,13 @@ free_anchor/freeanchor_r50_fpn_1x_coco.py
 fsaf/fsaf_r50_fpn_1x_coco.py
 gcnet/mask-rcnn_r50-gcb-r4-c3-c5_fpn_1x_coco.py
 gfl/gfl_r50_fpn_1x_coco.py
+glip/glip_atss_swin-t_a_fpn_dyhead_pretrain_obj365.py
 ghm/retinanet_r50_fpn_ghm-1x_coco.py
 gn/mask-rcnn_r50_fpn_gn-all_2x_coco.py
 gn+ws/faster-rcnn_r50_fpn_gn-ws-all_1x_coco.py
-grid_rcnn/grid-rcnn_r50_fpn_gn-head_1x_coco.py
+grid_rcnn/grid-rcnn_r50_fpn_gn-head_2x_coco.py
 groie/faste-rcnn_r50_fpn_groie_1x_coco.py
-guided_anchoring/ga-faster-rcnn_r50_fpn_1x_coco.py
+guided_anchoring/ga-faster-rcnn_r50-caffe_fpn_1x_coco.py
 hrnet/htc_hrnetv2p-w18_20e_coco.py
 htc/htc_r50_fpn_1x_coco.py
 instaboost/mask-rcnn_r50_fpn_instaboost-4x_coco.py
@@ -50,7 +51,7 @@ mask2former/mask2former_r50_8xb2-lsj-50e_coco.py
 mask2former/mask2former_r50_8xb2-lsj-50e_coco-panoptic.py
 mask_rcnn/mask-rcnn_r50_fpn_1x_coco.py
 maskformer/maskformer_r50_ms-16xb1-75e_coco.py
-ms_rcnn/ms-rcnn_r50_fpn_1x_coco.py
+ms_rcnn/ms-rcnn_r50-caffe_fpn_1x_coco.py
 nas_fcos/nas-fcos_r50-caffe_fpn_nashead-gn-head_4xb4-1x_coco.py
 nas_fpn/retinanet_r50_nasfpn_crop640-50e_coco.py
 paa/paa_r50_fpn_1x_coco.py
@@ -79,13 +80,17 @@ solo/solo_r50_fpn_1x_coco.py
 solov2/solov2_r50_fpn_1x_coco.py
 sparse_rcnn/sparse-rcnn_r50_fpn_1x_coco.py
 ssd/ssd300_coco.py
-strong_baselines/mask-rcnn_r50-caffe_fpn_rpn-2conv_4conv1fc_syncbn-all_amp-lsj-100e_coco.py
 swin/mask-rcnn_swin-t-p4-w7_fpn_1x_coco.py
-timm_example/retinanet_timm-tv-resnet50_fpn_1x_coco.py
 tood/tood_r50_fpn_1x_coco.py
 tridentnet/tridentnet_r50-caffe_1x_coco.py
 vfnet/vfnet_r50_fpn_1x_coco.py
 yolact/yolact_r50_8xb8-55e_coco.py
 yolo/yolov3_d53_8xb8-320-273e_coco.py
 yolof/yolof_r50-c5_8xb8-1x_coco.py
 yolox/yolox_s_8xb8-300e_coco.py
+deepsort/deepsort_faster-rcnn_r50_fpn_8xb2-4e_mot17halftrain_test-mot17halfval.py
+mask2former_vis/mask2former_r50_8xb2-8e_youtubevis2021.py
+masktrack_rcnn/masktrack-rcnn_mask-rcnn_r50_fpn_8xb1-12e_youtubevis2021.py
+ocsort/ocsort_yolox_x_8xb4-amp-80e_crowdhuman-mot17halftrain_test-mot17halfval.py
+qdtrack/qdtrack_faster-rcnn_r50_fpn_8xb2-4e_mot17halftrain_test-mot17halfval.py
+strongsort/strongsort_yolox_x_8xb4-80e_crowdhuman-mot17halftrain_test-mot17halfval.py

Diff for: .dev_scripts/benchmark_options.py

@@ -5,8 +5,11 @@
     'pip install instaboostfast',
     'pip install git+https://github.com/cocodataset/panopticapi.git',
     'pip install timm',
-    'pip install mmcls>=1.0.0rc0',
+    'pip install mmpretrain',
     'pip install git+https://github.com/lvis-dataset/lvis-api.git',
+    'pip install -r ../requirements/multimodal.txt',
+    'pip install -r ../requirements/tracking.txt',
+    'pip install git+https://github.com/JonathonLuiten/TrackEval.git',
 ]
 
 default_floating_range = 0.5

Diff for: .dev_scripts/benchmark_train_models.txt

@@ -15,4 +15,6 @@ mask2former/mask2former_r50_8xb2-lsj-50e_coco-panoptic.py
 swin/mask-rcnn_swin-t-p4-w7_fpn_1x_coco.py
 condinst/condinst_r50_fpn_ms-poly-90k_coco_instance.py
 lvis/mask-rcnn_r50_fpn_sample1e-3_ms-1x_lvis-v1.py
-convnext/mask-rcnn_convnext-t-p4-w7_fpn_amp-ms-crop-3x_coco.py
+mask2former_vis/mask2former_r50_8xb2-8e_youtubevis2021.py
+masktrack_rcnn/masktrack-rcnn_mask-rcnn_r50_fpn_8xb1-12e_youtubevis2021.py
+qdtrack/qdtrack_faster-rcnn_r50_fpn_8xb2-4e_mot17halftrain_test-mot17halfval.py

Diff for: .dev_scripts/gather_models.py

@@ -1,10 +1,11 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import argparse
 import glob
-import json
+import os
 import os.path as osp
 import shutil
 import subprocess
+import time
 from collections import OrderedDict
 
 import torch
@@ -32,11 +33,15 @@ def process_checkpoint(in_file, out_file):
     # remove optimizer for smaller file size
     if 'optimizer' in checkpoint:
         del checkpoint['optimizer']
+    if 'ema_state_dict' in checkpoint:
+        del checkpoint['ema_state_dict']
 
     # remove ema state_dict
     for key in list(checkpoint['state_dict']):
         if key.startswith('ema_'):
             checkpoint['state_dict'].pop(key)
+        elif key.startswith('data_preprocessor'):
+            checkpoint['state_dict'].pop(key)
 
     # if it is necessary to remove some sensitive data in checkpoint['meta'],
     # add the code here.
@@ -52,15 +57,15 @@ def process_checkpoint(in_file, out_file):
 
 def is_by_epoch(config):
     cfg = Config.fromfile('./configs/' + config)
-    return cfg.runner.type == 'EpochBasedRunner'
+    return cfg.train_cfg.type == 'EpochBasedTrainLoop'
 
 
 def get_final_epoch_or_iter(config):
     cfg = Config.fromfile('./configs/' + config)
-    if cfg.runner.type == 'EpochBasedRunner':
-        return cfg.runner.max_epochs
+    if cfg.train_cfg.type == 'EpochBasedTrainLoop':
+        return cfg.train_cfg.max_epochs
     else:
-        return cfg.runner.max_iters
+        return cfg.train_cfg.max_iters
 
 
 def get_best_epoch_or_iter(exp_dir):
@@ -74,60 +79,22 @@ def get_best_epoch_or_iter(exp_dir):
 
 def get_real_epoch_or_iter(config):
     cfg = Config.fromfile('./configs/' + config)
-    if cfg.runner.type == 'EpochBasedRunner':
-        epoch = cfg.runner.max_epochs
-        if cfg.data.train.type == 'RepeatDataset':
-            epoch *= cfg.data.train.times
+    if cfg.train_cfg.type == 'EpochBasedTrainLoop':
+        epoch = cfg.train_cfg.max_epochs
         return epoch
     else:
-        return cfg.runner.max_iters
+        return cfg.train_cfg.max_iters
 
 
 def get_final_results(log_json_path,
                       epoch_or_iter,
-                      results_lut,
+                      results_lut='coco/bbox_mAP',
                       by_epoch=True):
     result_dict = dict()
-    last_val_line = None
-    last_train_line = None
-    last_val_line_idx = -1
-    last_train_line_idx = -1
-    with open(log_json_path, 'r') as f:
-        for i, line in enumerate(f.readlines()):
-            log_line = json.loads(line)
-            if 'mode' not in log_line.keys():
-                continue
-
-            if by_epoch:
-                if (log_line['mode'] == 'train'
-                        and log_line['epoch'] == epoch_or_iter):
-                    result_dict['memory'] = log_line['memory']
-
-                if (log_line['mode'] == 'val'
-                        and log_line['epoch'] == epoch_or_iter):
-                    result_dict.update({
-                        key: log_line[key]
-                        for key in results_lut if key in log_line
-                    })
-                    return result_dict
-            else:
-                if log_line['mode'] == 'train':
-                    last_train_line_idx = i
-                    last_train_line = log_line
-
-                if log_line and log_line['mode'] == 'val':
-                    last_val_line_idx = i
-                    last_val_line = log_line
-
-    # bug: max_iters = 768, last_train_line['iter'] = 750
-    assert last_val_line_idx == last_train_line_idx + 1, \
-        'Log file is incomplete'
-    result_dict['memory'] = last_train_line['memory']
-    result_dict.update({
-        key: last_val_line[key]
-        for key in results_lut if key in last_val_line
-    })
-
+    with open(log_json_path) as f:
+        r = f.readlines()[-1]
+        last_metric = r.split(',')[0].split(': ')[-1].strip()
+    result_dict[results_lut] = last_metric
     return result_dict
 
 
@@ -150,6 +117,16 @@ def get_dataset_name(config):
     return name_map[cfg.dataset_type]
 
 
+def find_last_dir(model_dir):
+    dst_times = []
+    for time_stamp in os.scandir(model_dir):
+        if osp.isdir(time_stamp):
+            dst_time = time.mktime(
+                time.strptime(time_stamp.name, '%Y%m%d_%H%M%S'))
+            dst_times.append([dst_time, time_stamp.name])
+    return max(dst_times, key=lambda x: x[0])[1]
+
+
 def convert_model_info_to_pwc(model_infos):
     pwc_files = {}
     for model in model_infos:
@@ -160,9 +137,7 @@ def convert_model_info_to_pwc(model_infos):
         pwc_model_info['Config'] = osp.join('configs', model['config'])
 
         # get metadata
-        memory = round(model['results']['memory'] / 1024, 1)
         meta_data = OrderedDict()
-        meta_data['Training Memory (GB)'] = memory
         if 'epochs' in model:
             meta_data['Epochs'] = get_real_epoch_or_iter(model['config'])
         else:
@@ -198,7 +173,7 @@ def convert_model_info_to_pwc(model_infos):
                     Metrics={'PQ': metric}))
         pwc_model_info['Results'] = results
 
-        link_string = 'https://download.openmmlab.com/mmdetection/v2.0/'
+        link_string = 'https://download.openmmlab.com/mmdetection/v3.0/'
         link_string += '{}/{}'.format(model['config'].rstrip('.py'),
                                       osp.split(model['model_path'])[-1])
         pwc_model_info['Weights'] = link_string
@@ -214,9 +189,13 @@ def parse_args():
     parser.add_argument(
         'root',
         type=str,
+        default='work_dirs',
         help='root path of benchmarked models to be gathered')
     parser.add_argument(
-        'out', type=str, help='output path of gathered models to be stored')
+        '--out',
+        type=str,
+        default='gather',
+        help='output path of gathered models to be stored')
     parser.add_argument(
         '--best',
         action='store_true',
@@ -262,32 +241,22 @@ def main():
             continue
 
         # get the latest logs
-        log_json_path = list(
-            sorted(glob.glob(osp.join(exp_dir, '*.log.json'))))[-1]
-        log_txt_path = list(sorted(glob.glob(osp.join(exp_dir, '*.log'))))[-1]
-        cfg = Config.fromfile('./configs/' + used_config)
-        results_lut = cfg.evaluation.metric
-        if not isinstance(results_lut, list):
-            results_lut = [results_lut]
-        # case when using VOC, the evaluation key is only 'mAP'
-        # when using Panoptic Dataset, the evaluation key is 'PQ'.
-        for i, key in enumerate(results_lut):
-            if 'mAP' not in key and 'PQ' not in key:
-                results_lut[i] = key + '_mAP'
-        model_performance = get_final_results(log_json_path,
-                                              final_epoch_or_iter, results_lut,
-                                              by_epoch)
+        latest_exp_name = find_last_dir(exp_dir)
+        latest_exp_json = osp.join(exp_dir, latest_exp_name, 'vis_data',
+                                   latest_exp_name + '.json')
+
+        model_performance = get_final_results(
+            latest_exp_json, final_epoch_or_iter, by_epoch=by_epoch)
 
         if model_performance is None:
             continue
 
-        model_time = osp.split(log_txt_path)[-1].split('.')[0]
         model_info = dict(
             config=used_config,
             results=model_performance,
-            model_time=model_time,
             final_model=final_model,
-            log_json_path=osp.split(log_json_path)[-1])
+            latest_exp_json=latest_exp_json,
+            latest_exp_name=latest_exp_name)
         model_info['epochs' if by_epoch else 'iterations'] =\
             final_epoch_or_iter
         model_infos.append(model_info)
@@ -300,7 +269,7 @@ def main():
 
         model_name = osp.split(model['config'])[-1].split('.')[0]
 
-        model_name += '_' + model['model_time']
+        model_name += '_' + model['latest_exp_name']
         publish_model_path = osp.join(model_publish_dir, model_name)
         trained_model_path = osp.join(models_root, model['config'],
                                       model['final_model'])
@@ -310,13 +279,8 @@ def main():
                                               publish_model_path)
 
         # copy log
-        shutil.copy(
-            osp.join(models_root, model['config'], model['log_json_path']),
-            osp.join(model_publish_dir, f'{model_name}.log.json'))
-        shutil.copy(
-            osp.join(models_root, model['config'],
-                     model['log_json_path'].rstrip('.json')),
-            osp.join(model_publish_dir, f'{model_name}.log'))
+        shutil.copy(model['latest_exp_json'],
+                    osp.join(model_publish_dir, f'{model_name}.log.json'))
 
         # copy config to guarantee reproducibility
         config_path = model['config']

Diff for: .github/pull_request_template.md

@@ -21,5 +21,5 @@ If this PR introduces a new feature, it is better to list some use cases here, a
 
 1. Pre-commit or other linting tools are used to fix the potential lint issues.
 2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
-3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMCls.
+3. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMPreTrain.
 4. The documentation has been modified accordingly, like docstring or example tutorials.

Diff for: .readthedocs.yml

@@ -1,9 +1,14 @@
 version: 2
 
-formats: all
+build:
+  os: ubuntu-22.04
+  tools:
+    python: "3.8"
+
+formats:
+  - epub
 
 python:
-  version: 3.7
   install:
     - requirements: requirements/docs.txt
     - requirements: requirements/readthedocs.txt