RobMulla / chessboard-vision

Chessboard Computer Vision Project

chessboard-vision

Chessboard Computer Vision Project

Project Overview

I would like to create a dataset of annotated chess board images with the board setup. Then I would like to train a deep learning model to classify chess board positions using nothing but images. This may be a good dataset for a community kaggle competition (pogchamps?)

Eventually the project would produce an app that could be used in real time to track a game's progress using video.

Workflow

0. Read existing papers on the topic.

• https://web.stanford.edu/class/cs231a/prev_projects_2016/CS_231A_Final_Report.pdf
• http://vision.fe.uni-lj.si/cvww2016/proceedings/papers/21.pdf

1. Gather the data

• Youtube scraping
  • Need to hand label or extract from game labels.
• My own images- setup the board and label.

Things to figure out

1. How the labels will look?
   • FEN
2. Test out object detection

Reference:

• https://github.com/Paethon/chess_image_dataset <- 3d image creation of chess boards?

TODO (March 22, 2022)

1. Automatically identify X,Y limits of the GT board.
2. Crop to the IRL board.

TODO (April 3, 2022)