VisionPlay : Next-Gen Sports Analytics
Developed an advanced system using computer vision and machine learning to analyze soccer match footage, tracking players,referees, and the ball in real-time to provide insights into performance and strategies.
Computer Vision
1 month
Multi-Object Detection:
Fine-tuned a YOLO model for a custom dataset of players, goalkeepers, and the ball, ensuring accurate real-time detection and utilizing ByteTrack for robust tracking.
Player Identification:
Automated team assignment using K-means clustering and uniquely tracked individual players.
Ball Possession Analysis:
Determined ball possession per frame and displayed real-time team control statistics.
Performance Metrics:
Computed player speed and distance covered, displaying metrics in real-time.
Camera Compensation:
Estimated and compensated for camera movement using optical flow techniques.
Perspective Transformation:
Mapped pixel coordinates to actual field positions for spatial analysis.
Visual Annotations:
Overlaid bounding boxes and performance metrics on video footage.
Data Processing:
Implemented batch processing for efficient handling of large video files.
