Player Re‑Identification & Tracking

This repository provides a simple pipeline to detect and re‑identify (track) players in a sports video using:

• YOLO (Ultralytics) for player detection
• ViT (Timm) for appearance feature extraction
• Kalman Filter + Hungarian assignment + fused cost (IoU, appearance similarity, center distance) for robust tracking

---

Contents

• player_identify.py — main tracking script
• requirements.txt — Python dependencies
• Example video/model paths configured inside player_identify.py

---

Quick Start

1. Clone this repository
2. Install Python 3.13 and create a virtual environment

   python3 -m venv .venv
   source .venv/bin/activate      # Linux / macOS
   .\.venv\Scripts\activate       # Windows PowerShell

Adjust the model_path (MODEL_PATH) , Input video path (VIDEO_PATH) and output video path (OUTPUT_PATH) inside player_identify.py