🎮 Mario Fixed-ID Tracker

This project provides a custom object tracking system using YOLOv8 and ByteTrack to analyze gameplay videos.
It focuses on tracking specific game characters with fixed IDs — including Luigi, Boo, Peach, and Shy Guy — and dynamically tracks generic obstacles using ByteTrack.

---

🔧 Features

• 🎯 Fixed ID tracking for main characters:
  • Luigi → ID 1
  • Boo → ID 2
  • Peach → ID 3
  • Shy Guy → ID 4
• 🟩 Dynamic object tracking for 'obstacle' using ByteTrack
• ✅ Works with YOLOv8 fine-tuned custom models
• 🕹 Accurate for gameplay analysis, emotion interaction, or data collection

---

📁 Folder Structure

/fine-tune
├── test/
│   └── Gameplay.mp4               # Input video
├── output_tracked/
│   └── 20240515-conf0.4.mp4       # Output video with tracked boxes
├── runs/
│   └── detect/train14/weights/    # Your YOLOv8 trained model
├── yolov8_fixedid_bytetrack.py    # Main tracking script

---

🚀 Quick Start

1. Install dependencies

pip install -r requirements.txt

2. Place your fine-tuned YOLOv8 model at:

runs/detect/trainX/weights/best.pt

3. Put your input video here:

./test/Gameplay.mp4

4. Run the tracker

cd fine-tune/
python yolov8_fixedid_bytetrack.py

5. Output will be saved as:

./output_tracked/<timestamp>-conf<score>.mp4

---

🧠 Model Requirements

Make sure your YOLOv8 model was trained with the following class names in data.yaml:

names:
  0: Luigi
  1: Boo
  2: Peach
  3: Shy Guy
  4: obstacle

---

📈 Example Output

---

📤 Future Ideas

• CSV export of tracking logs (frame-by-frame position, ID)
• Invisible persistence (tracking through occlusion)
• Web UI to review tracking results
• Interactive labeling pipeline

---

📄 License

MIT License. This repo is designed for research and non-commercial use in gameplay studies and interaction design.