README.md

ChessApp

Deep Learning for computer vision project. Android application for chess board detection and pieces prediction using a neural network for classification trained on home-made dataset and tested on different techniques. The application also creates a virtual representation of the board allowing the user to visualize the classification results.

Datasets

For the dataset we created our own by collenting chess board images, crop each box and label them. We have a small dataset made of 12.500 images and an augmented one made by 110.600 images.
All the datasets can be found here.

Models

We generated up to 13 models using different approaches and training techniques.

13 classes approach

2 + 7 classes approach

2 classes:

7 classes:

All the checkpoints can be found here.