Releases: linrock/chessboard-recognizer
Releases · linrock/chessboard-recognizer
NN model for tensorflow 2.3.1
NN model trained on more grayscale images
- More chessboard training images
- 390 auto-generated images
- 126 manually-created based on classifier mistakes
Updated NN model with black-and-white PDF images
- Training images include more manually-verified black-and-white chessboard images from PDFs
- Updated NN model uses 29,248 training images
Trained model on more board types
- Updated nn-model to better detect Lichess images and book diagrams
- Added more training images
Initial neural network model and training images
- nn-model.zip contains a trained tensorflow 2 CNN model that can be used with recognize.py
- training-images.zip contains all images used to train this version of the neural network