This project implements a real-time human pose estimation system using YOLOv8 Pose model and OpenCV. The system captures video (webcam or uploaded file), detects humans, and identifies 17 key body landmarks (shoulders, elbows, knees, etc.) in real time. This demonstrates end-to-end integration of deep learning models into a live computer vision pipeline.
Python
YOLOv8 Pose (Ultralytics)
OpenCV
- Real-time video processing
- 17 human keypoint detection
- Skeleton visualization
- Lightweight and fast inference
- Easy-to-deploy architecture
Video Input → Frame Capture → YOLOv8 Pose Model → Keypoint Detection → Skeleton Rendering → Real-Time Display
pip install ultralytics opencv-python
python app.py
Press q to exit.
Capture frame using OpenCV
Pass frame to YOLOv8 pose model
Bounding boxes
Confidence scores
17 human keypoints
Skeleton connections are drawn
🏋️ Fitness posture correction
🧘 Yoga pose analysis
⚽ Sports performance tracking
🤖 Robotics motion understanding
🏢 Workplace ergonomic monitoring
Real-time inference (hardware dependent)
Works on CPU (basic performance)
Optimized performance with GPU