नियत ही नियति है |

Video Segmentation algorithms

The script implements color-based segmentation that identifies objects by their HSV color ranges, making it particularly effective for scenarios where you need to isolate blue, green, or red objects from the scene. Background subtraction using the MOG2 algorithm excels at detecting moving objects by learning what constitutes the static background over time. Edge detection combines three powerful methods – Canny, Sobel, and Laplacian – to create comprehensive boundary maps of objects and structures within each frame.

K-means clustering groups similar pixel colors into configurable clusters, essentially creating a posterized version of your video that can reveal underlying color patterns. The watershed algorithm treats pixel intensities as topographic elevations, finding natural boundaries between regions through a flooding simulation approach. Optical flow segmentation tracks motion between consecutive frames, highlighting areas of movement and creating dynamic motion maps.

GitHub Link: https://github.com/akshansh11/Video-Segmentation-Pipeline

Motion Tracking in a video file

I built an AI-powered motion tracking system that analyzes human movement in videos using computer vision. The project offers three tracking modes: a basic single-person tracker using MediaPipe for pose detection, a multi-person tracker that color-codes each individual, and an advanced YOLO-based system that handles crowded scenes with 10+ people.

The system calculates real-time metrics like movement velocity, joint angles, and energy levels, then exports everything to JSON for further analysis. It generates annotated videos with skeleton overlays and bounding boxes, plus detailed reports ranking participants by activity levels and screen time.

Built with Python, OpenCV, MediaPipe, and YOLOv8, it’s perfect for analyzing sports performance, dance choreography, fitness routines, or any scenario where tracking human movement matters. The code handles video validation, filters false detections, and maintains persistent tracking IDs across frames. Installation is straightforward with pip, and the modular design makes it easy to customize for specific use cases.

GitHub Repo: https://github.com/akshansh11/Motion-Movement-Tracker