Chess Scan

Detect chess positions from images and convert them to FEN strings.

Features

• Detects chessboards from screenshots and images
• Recognizes all chess pieces
• Generates valid FEN strings
• Supports local files, URLs, buffers, and raw image data

Installation

npm install chess-scan

Usage

Generate FEN

const { imageToFen } = require('chess-scan')

async function main() {
	const result = await imageToFen('board.png')

	console.log(result.fen)
}

main()

Extract Board Image

const { getBoardImg } = require('chess-scan')

async function main() {
	const board = await getBoardImg('board.png')

	console.log(board.width)
	console.log(board.height)
}

Extract Individual Squares

const { getSquareImgs } = require('chess-scan')

async function main() {
	const squares = await getSquareImgs('board.png')

	console.log(squares.length) // 64
}

API

imageToFen(input)

Detects the board, recognizes all pieces, and returns the resulting FEN.

imageToFen(input: ImageInput): Promise<{
        fen: string
        confusion: [{
            classId: number
            confidence: number
        },...]
}>

getBoardImg(input)

Returns the detected chessboard as a normalized image.

getBoardImg(input: ImageInput): Promise<Buffer>

getSquareImgs(input)

Returns all 64 board squares.

getSquareImgs(input: ImageInput): Promise<Buffer[]>

Supported Inputs

The library accepts:

• Local file paths
• Image URLs
• Buffers
• Raw image data

TypeScript

Type definitions are included automatically.

import { imageToFen } from 'chess-scan'

Pipeline

The library uses a machine-learning pipeline consisting of:

1. Board detection
2. Board normalization
3. Square extraction
4. Piece classification
5. FEN generation

Performance depends on image quality, board visibility, and piece style.

License

MIT