C++ DNN backend, from https://github.com/sony/nnabla (for train) https://github.com/davisking/dlib (for player, forward only)

mcts & go game engine, from https://github.com/gcp/leela-zero https://github.com/ymgaq/AQ

DarknetGo CNN and trained weights, from https://github.com/pjreddie/darknet

Board UI, res from https://github.com/waltheri/wgo.js ui engine from https://github.com/davisking/dlib

Q: Why implementing CNN using different NN backend? A: nnabla is too big for me. I want my nn player engine small and easy to embedding.