Skip to content

bahelms/chess_vision

Repository files navigation

Chess Vision

Convert an image of a chessboard into a digital representation

Battle Plan

  • LiveView to capture image and store it. File or S3?
  • Parse image into array of images of 64 squares
    • Canny edge detection - Rust nif
    • Hough line detection - Rust nif
    • order is important: top left of board will be element 0
      • orientation is not guaranteed; A1 could be any corner of the board
  • Convert each of those images into data points by feeding them to an image recognition neural network
    • train NN
      • training/test sets of tactics book images and live board images. Use step 1 to help with this.
      • Use Nx
    • output is array of structs containing { piece_type, piece_color, square_color }
  • Convert that data into a FEN string
  • Render FEN string as a chess board in the web app UI

About

Chessboard image recognition

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published