Welcome to the Naruto Hand Sign Trainer!
This project is a full system combining:
- 📚 A Training Notebook to prepare and augment a dataset of Naruto hand signs.
- 🖥️ A Flask WebApp that detects your hand signs in real-time, teaching you jutsus like Shadow Clone Jutsu, Chidori, and more!
Built with ❤️ using Flask, YOLOv8, Mediapipe, OpenCV, and custom augmentation techniques.
-
📂 Load and process a custom Naruto hand sign dataset.
-
🔄 Data Augmentation (horizontal flipping, scaling, HSV shifts).
-
🏋️ Fine-tune a YOLOv8 model (
best_naruto_signs_7.pt) for hand sign classification. -
📈 Train with optimized parameters for small datasets.
Here’s the full training pipeline from data to deployment:
)
✅ Load Dataset
✅ Augment Data (Flip, Color, Scale)
✅ Format for YOLOv8
✅ Configure and Train Model
✅ Evaluate and Export Best Model
✅ Deploy to Flask WebApp
- 🖐️ Real-time Hand Sign Detection using YOLOv8 and Mediapipe.
- 📚 Learning Mode: Follow jutsu sequences step-by-step.
- 🎯 Mastery Mode: Perform full jutsus flawlessly to level up.
- 🆓 Free Practice Mode: Practice any sign freely.
- ⏱️ Time Attack Mode: Complete as many hand signs as you can in 60 seconds.
- 📝 Gesture history saved into
gestures.json. - 📈 Mastery tracking saved into
stats.json.
/app/
├── main.py # Flask app for real-time sign detection
├── gestures.json # Saved detected hand signs
├── stats.json # Jutsu performance tracking
├── templates/
│ ├── menu.html # Main menu page
│ ├── index.html # Live detection page
│ ├── stats.html # Statistics page
/notebooks/
├── codeitup5-0.ipynb # Training and Augmentation Notebook
/models/
├── best_naruto_signs_7.pt # Trained YOLO model (not included in repo)
requirements.txt # Project dependencies
README.md # This file!