board-to-fen 0.1.1

An package that converts digial chessboard image into Forsyth-Edwards notation

board_to_fen

Python package that converts digital chessboard image into Forsyth-Edwards notation (FEN) notation

Installation

board_to_fen is available on PyPI:

$ pip3 install board_to_fen

Quick Start

from board_to_fen.predict import get_fen_from_image

print(get_fen_from_image(PATH_TO_CHESSBOARD_IMAGE))

Note: The package uses tensorflow+keras and python-opencv API. They are pretty heavy.

Customization

get_fen_from_image takes has 3 arguments:

• image_path [required]
• end_of_row '/' by default
• black_view False by default -> set True if chessboard is provided from black player perspective

Web version (currently may not work)

Available at: https://board2fen.bieda.it

Training

For training You would probably want to download the source code by cloning the repository:

$ git clone https://github.com/mcdominik/board_to_fen.git

Download training data from:
I will supply url for data in the future

In the main repository dir, run

$ python3 ./board_to_fen/train_model.py

Version history

• january 2023
  • versions 0.0.17-25
  • added simple board validation
  • bug fixes

Warnings

• Image has to be provided in neutral angle (white or black player's perspective).
• Image has to be square (~3% tolerance depending on image resolution).
• Image can't contain paddings, board borders etc. other than 64 squares (with pieces) itself.

References:

https://www.kaggle.com/datasets/koryakinp/chess-positions