Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -1,68 +1,49 @@
|
|
| 1 |
-
|
| 2 |
-
language: en
|
| 3 |
-
tags:
|
| 4 |
-
- computer-vision
|
| 5 |
-
- object-detection
|
| 6 |
-
- yolo
|
| 7 |
-
- chess
|
| 8 |
-
- ultralytics
|
| 9 |
-
license: mit
|
| 10 |
-
---
|
| 11 |
|
| 12 |
-
|
| 13 |
|
| 14 |
-
|
| 15 |
|
| 16 |
-
##
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
-
###
|
| 19 |
-
-
|
| 20 |
-
-
|
| 21 |
-
-
|
|
|
|
|
|
|
| 22 |
|
| 23 |
-
###
|
| 24 |
-
-
|
| 25 |
-
- **
|
| 26 |
-
-
|
|
|
|
|
|
|
| 27 |
|
|
|
|
| 28 |
|
| 29 |
-
### 🌊 Yonko v1.0 (`yonko_v1.0.pt`)
|
| 30 |
-
- **Usage**: Modèle entraîné sur un large dataset (10000 images) avec augmentation de données
|
| 31 |
-
- **Classes**: Toutes les pièces d'échecs
|
| 32 |
-
- **Dataset**: 10000+ images avec augmentation
|
| 33 |
-
- **Recommandé pour**: Détection robuste avec grande variété de conditions
|
| 34 |
|
| 35 |
-
## Utilisation
|
| 36 |
|
| 37 |
-
|
| 38 |
-
from huggingface_hub import hf_hub_download
|
| 39 |
-
from ultralytics import YOLO
|
| 40 |
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
)
|
| 47 |
-
model = YOLO(model_path)
|
| 48 |
|
| 49 |
-
#
|
| 50 |
-
|
| 51 |
-
|
| 52 |
|
| 53 |
-
##
|
|
|
|
| 54 |
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
Repository: https://github.com/MedouneSGB/Senchess-AI
|
| 58 |
-
|
| 59 |
-
## Modèles uploadés
|
| 60 |
-
|
| 61 |
-
- ✅ gear_v1.1.pt
|
| 62 |
-
- ✅ haki_v1.0.pt
|
| 63 |
-
|
| 64 |
-
- ✅ yonko_v1.0.pt
|
| 65 |
-
|
| 66 |
-
## License
|
| 67 |
-
|
| 68 |
-
MIT License - Libre d'utilisation
|
|
|
|
| 1 |
+
# Senchess AI Models
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
|
| 3 |
+
Collection of YOLOv8 models for chess piece detection.
|
| 4 |
|
| 5 |
+
## 📦 Available Models
|
| 6 |
|
| 7 |
+
### 🐉 Kaido v1.0 (kaido_v1.0.pt) - **RECOMMENDED**
|
| 8 |
+
- **Best overall performance** on styled and standard boards
|
| 9 |
+
- **32/32 pieces detection** with 98.61% confidence
|
| 10 |
+
- Trained on 1,000 positions, 50 epochs
|
| 11 |
+
- **mAP50: 99.95%** | Precision: 99.5% | Recall: 99.5%
|
| 12 |
+
- Size: 6.25 MB
|
| 13 |
+
- Use case: General purpose, excellent for all board styles
|
| 14 |
|
| 15 |
+
### 🔱 Haki v1.0 (haki_v1.0.pt)
|
| 16 |
+
- Specialized for strategic pieces (King, Queen, Rook, Bishop)
|
| 17 |
+
- 32/32 pieces detection with 98.52% confidence
|
| 18 |
+
- Trained on styled boards
|
| 19 |
+
- Size: 6.25 MB
|
| 20 |
+
- Use case: Styled/artistic chess boards
|
| 21 |
|
| 22 |
+
### ⚙️ Gear v1.1 (gear_v1.1.pt)
|
| 23 |
+
- Optimized for standard wooden boards
|
| 24 |
+
- **mAP50: 98.79%**
|
| 25 |
+
- Size: 6.25 MB
|
| 26 |
+
- Use case: Classic chess boards only
|
| 27 |
+
- ⚠️ May fail on styled boards
|
| 28 |
|
| 29 |
+
## 🚀 Usage
|
| 30 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
|
|
|
|
| 32 |
|
| 33 |
+
## 📊 Performance Comparison
|
|
|
|
|
|
|
| 34 |
|
| 35 |
+
| Model | Pieces Detected | Confidence | mAP50 | Best For |
|
| 36 |
+
|-------|----------------|------------|-------|----------|
|
| 37 |
+
| 🐉 Kaido v1.0 | 32/32 | 98.61% | 99.95% | All boards ⭐ |
|
| 38 |
+
| 🔱 Haki v1.0 | 32/32 | 98.52% | - | Styled boards |
|
| 39 |
+
| ⚙️ Gear v1.1 | 0-32 | varies | 98.79% | Standard boards |
|
|
|
|
|
|
|
| 40 |
|
| 41 |
+
## 🎯 Classes (12)
|
| 42 |
+
- white-king, white-queen, white-rook, white-bishop, white-knight, white-pawn
|
| 43 |
+
- black-king, black-queen, black-rook, black-bishop, black-knight, black-pawn
|
| 44 |
|
| 45 |
+
## 📝 License
|
| 46 |
+
MIT
|
| 47 |
|
| 48 |
+
## �� Repository
|
| 49 |
+
https://github.com/MedouneSGB/Senchess-AI
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|