src/processing/imagePreprocessing.js

/**
 * Image Preprocessing - Prepare images for model inference
 */

import * as ort from 'https://cdn.jsdelivr.net/npm/onnxruntime-web@1.17.0/dist/esm/ort.min.js';
import { DETECTION_CONFIG } from '../config.js';
import { loadImage } from '../utils/imageUtils.js';

/**
 * Preprocess image for SigLIP models
 */
export async function preprocessImage(imageData, size) {
    const img = await loadImage(imageData);
    
    // Create canvas and resize
    const canvas = document.createElement('canvas');
    canvas.width = size;
    canvas.height = size;
    const ctx = canvas.getContext('2d');
    ctx.drawImage(img, 0, 0, size, size);
    
    // Get image data
    const imageDataObj = ctx.getImageData(0, 0, size, size);
    const data = imageDataObj.data;
    
    // Convert to float32 tensor [1, 3, size, size] and normalize
    const float32Data = new Float32Array(3 * size * size);
    const { mean, std } = DETECTION_CONFIG.siglip;
    
    for (let i = 0; i < size * size; i++) {
        float32Data[i] = ((data[i * 4] / 255.0) - mean[0]) / std[0]; // R
        float32Data[size * size + i] = ((data[i * 4 + 1] / 255.0) - mean[1]) / std[1]; // G
        float32Data[2 * size * size + i] = ((data[i * 4 + 2] / 255.0) - mean[2]) / std[2]; // B
    }
    
    return new ort.Tensor('float32', float32Data, [1, 3, size, size]);
}

/**
 * Preprocess image for YOLO
 */
export async function preprocessImageYOLO(img, size) {
    const canvas = document.createElement('canvas');
    canvas.width = size;
    canvas.height = size;
    const ctx = canvas.getContext('2d');
    ctx.drawImage(img, 0, 0, size, size);
    
    const imageData = ctx.getImageData(0, 0, size, size);
    const data = imageData.data;
    
    const float32Data = new Float32Array(3 * size * size);
    for (let i = 0; i < size * size; i++) {
        float32Data[i] = data[i * 4] / 255.0;
        float32Data[size * size + i] = data[i * 4 + 1] / 255.0;
        float32Data[2 * size * size + i] = data[i * 4 + 2] / 255.0;
    }
    
    return new ort.Tensor('float32', float32Data, [1, 3, size, size]);
}