S1M0N38/chess-cv-chessvision

ChessVision Dataset (Piece Classification)

This dataset is derived from the ChessVision-3LC dataset by Gudbrand Tandberg, adapted for chess piece classification.

Dataset Description

This dataset contains images of chess pieces for classification. It is derived from the original ChessVision-3LC dataset and has the following modifications:

• Purpose: The dataset is intended for training and testing image classification models to recognize chess pieces.
• Structure: The images are organized in an ImageFolder structure, ready to be used with popular deep learning frameworks. The directories are named after the piece they contain (e.g., wP for white pawn, bK for black king, xx for empty squares).
• Image Format: All images are 32x32 pixel PNG files (RGB format), optimized for storage and fast loading.

For more information about the original data collection process, please refer to the original repository.

Categories

The dataset contains 13 categories:

• White pieces: wR (Rook), wN (Knight), wB (Bishop), wQ (Queen), wK (King), wP (Pawn)
• Black pieces: bR (Rook), bN (Knight), bB (Bishop), bQ (Queen), bK (King), bP (Pawn)
• Empty square: xx (no piece on the square)

Dataset Structure

The dataset is split into two subsets:

• Train: Training images
• Validation: Validation images

Each subset contains images organized by category in subdirectories.