What this repo is meant for: if you would to build your own chesspiece identification models, or if you want to deploy a Chess ID server.
First, grab the data: https://www.dropbox.com/s/618l4ddoykotmru/Chess%20ID%20Public%20Data.zip?dl=0
My experiments are available in the Jupyter notebook.
First, grab the Caffe model: https://www.dropbox.com/s/fmlt5ook8ugovid/finetune_chess_iter_5554.caffemodel?dl=0
Install all dependencies (Caffe, OpenCV, etc.). Edit paths in server.py and then run it.