Product Development Project - Heureka Robot Machine Vision

This repository provides a machine vision implementation for detecting human motions in real-time.

Current Features

• YOLOv8 for Person Detection: Utilizes the YOLOv8 model to detect humans in real-time.
• Motion Detection: Detects moving gestures by analyzing motion within the bounding box of detected persons.
• Flexible Input:
  • Process live input from a webcam camera.
  • Process video files from a folder.
  • Process live input from an Intel RealSense camera.

Setup

Hardware

• A computer with a webcam camera.
• OR an Intel RealSense camera.

Software

• Python 3.10+

Python Dependencies

After cloning, please install the required Python packages by running this command:

pip install -r requirements.txt

Dependencies currently include:

• opencv-python
• ultralytics
• pyrealsense2
• numpy

Usage

Currently, the script supports the following use cases:

1. Gesture detection from Intel RealSense camera or Webcam Camera

Run the following command:

python mediapipe_models/hand_gesture_recognizer.py

2. Motion detection from test videos

To process a folder of test videos (that are used as input to the model), follow these steps:

1. Place your video files in a folder (here it's test-videos).
2. Run the following command:

python with_stock_vid/main.py

The script will iterate through all video files in the specified folder and display the processed frames with motion detection results. .

Main Structure of Repository

Currently, the repository structure is as follows:

├── models/                         # For classes that use machine learning models
│   ├── motion_detector.py          # Contains the MotionDetector class
│   ├── yolov8n.py                  # The YOLOv8 model for object detection
├── with_stock_vid/                 
│   ├── main.py                     # Main script with stock videos
│   ├── test-videos/                # Folder containing test video files     
├── with_camera/                    
│   ├── main.py                     # Script to run motion detection from camera
├── requirements.txt                # Python dependencies
├── README.md                       # Documentation (this file)

Known Issues

• Motion Detection Sensitivity: Requires further fine-tuning of the motion detection threshold. The algorithm still needs to distinguish actual waving gesture to other motions.
• Deprecated Warnings: Some platforms may show warnings for deprecated camera APIs.
• Intel RealSense SDK Installation: Ensure pyrealsense2 is installed correctly for live camera input.

Acknowledgments

• YOLOv8 model by Ultralytics. Key features of YOLOv8 model here
• Intel RealSense SDK for Python