rambotanz2/yolo11n-video-soccer

Soccer Player Tracker

Upload a soccer match clip and the app will detect every player, give each one a stable colored ID, and keep that ID consistent across every frame.

How it works

1. 1.Detect — YOLO finds players in each frame
2. 2.Fingerprint — ResNet50 turns each player crop into a feature vector
3. 3.Track — IOU + appearance matching links the same player across frames
4. 4.Render — annotated MP4 with colored boxes + ID numbers per player

Run locally

git clone https://github.com/thunder500/yolo11n-video-soccer.git
cd yolo11n-video-soccer
python -m venv .venv
.venv/Scripts/activate          # Linux/Mac: source .venv/bin/activate
pip install -r requirements.txt
python app.py

Then open http://localhost:5000.

If you have a fine-tuned best.pt, drop it in the project root — otherwise the pipeline auto-downloads yolo11n.pt and filters to the COCO person class.

Run with Docker

docker build -t soccer-tracker .
docker run -p 7860:7860 soccer-tracker

Tech

YOLO · ResNet50 · OpenCV · PyTorch · Flask