End-to-end computer vision pipeline that

End-to-end computer vision pipeline that converts broadcast football video into 70+ tactical features. Led a 6-person team. A dual-detector system (YOLOv8 + RT-DETR second pass) pushed ball detection from 4.7% to 85.7%; ByteTrack handles identity continuity; jersey OCR resolves player identity. Runs at 25 FPS on consumer hardware across ~120K tracked frames. Delivered as a 16-tab Streamlit dashboard with xG modeling, formation detection, and pressing analysis.

Stack: Python, PyTorch, YOLOv8, RT-DETR, ByteTrack, OpenCV, Streamlit Link: github.com/MohanadddAmrr/Tactivision