Chess Move Tracking Models (YOLO11)

This repository contains fine-tuned models for a Chess Move Tracking pipeline.

Models

1. models/yolo11s_pose_chessboard.pt:

   • Architecture: YOLO11s-pose
   • Task: Keypoint Detection (Chessboard Corner Localization).
   • Classes: 1 class (chessboard).
   • Keypoints: a1, h1, a8, h8.
   • Input: Raw video frame (rotated/occluded).

2. models/yolo11m_pieces.pt:

   • Architecture: YOLO11m
   • Task: Object Detection (Chess Pieces Detection).
   • Classes: 13 classes (Hand, bB, bK, bN, bP, bQ, bR, wB, wK, wN, wP, wQ, wR).
   • Input: Warped top-down chessboard image (640x640).

Usage

You can use these models directly with the Ultralytics library.

1. Installation

First, install the required library:

pip install -U ultralytics

2. Using the Chessboard Corner Localization Model

This model detects the 4 semantic corners of the chessboard (a1, h1, a8, h8) to help with perspective warping.

from ultralytics import YOLO

# Load the model directly from Hugging Face
model = YOLO("https://huggingface.co/surawut/chess-move-tracking-yolo11/resolve/main/models/yolo11s_pose_chessboard.pt")
# Load an image (raw frame)
image_path = "path/to/raw_frame.jpg"
# Run inference
results = model(image_path)

3. Using the Chess Pieces Detection Model

This model detects the 12 chess pieces + hand on a warped (640x640) top-down view of the board.

from ultralytics import YOLO

# Load the model directly from Hugging Face
model = YOLO("https://huggingface.co/surawut/chess-move-tracking-yolo11/resolve/main/models/yolo11m_pieces.pt")
# Load a chessboard image
warped_image_path = "path/to/warped_board.jpg"
# Run inference
results = model(warped_image_path)