worker.js

import { pipeline, env } from 'https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.0.0-alpha.19';

// Memory & Performance Tuning
env.backends.onnx.wasm.numThreads = 2; 
env.useBrowserCache = true;

let objectDetector = null;
let targetLanguage = 'spa_Latn';
let detectionThreshold = 0.75;
let focusMode = true;

const labelTranslations = {
    spa_Latn: {
        person: 'persona',
        bicycle: 'bicicleta',
        car: 'coche',
        motorcycle: 'motocicleta',
        airplane: 'avion',
        bus: 'autobus',
        train: 'tren',
        truck: 'camion',
        boat: 'barco',
        'traffic light': 'semaforo',
        'fire hydrant': 'hidrante',
        'stop sign': 'senal de alto',
        bench: 'banco',
        bird: 'pajaro',
        cat: 'gato',
        dog: 'perro',
        horse: 'caballo',
        sheep: 'oveja',
        cow: 'vaca',
        backpack: 'mochila',
        umbrella: 'paraguas',
        handbag: 'bolso',
        tie: 'corbata',
        suitcase: 'maleta',
        frisbee: 'frisbee',
        skis: 'esquis',
        snowboard: 'tabla de snowboard',
        'sports ball': 'pelota',
        kite: 'cometa',
        'baseball bat': 'bate de beisbol',
        skateboard: 'patineta',
        surfboard: 'tabla de surf',
        bottle: 'botella',
        'wine glass': 'copa',
        cup: 'taza',
        fork: 'tenedor',
        knife: 'cuchillo',
        spoon: 'cuchara',
        bowl: 'tazon',
        banana: 'platano',
        apple: 'manzana',
        sandwich: 'sandwich',
        orange: 'naranja',
        broccoli: 'brocoli',
        carrot: 'zanahoria',
        pizza: 'pizza',
        donut: 'dona',
        cake: 'pastel',
        chair: 'silla',
        couch: 'sofa',
        bed: 'cama',
        table: 'mesa',
        toilet: 'inodoro',
        tv: 'televisor',
        laptop: 'portatil',
        mouse: 'raton',
        remote: 'control remoto',
        keyboard: 'teclado',
        'cell phone': 'telefono movil',
        microwave: 'microondas',
        oven: 'horno',
        toaster: 'tostadora',
        sink: 'fregadero',
        refrigerator: 'refrigerador',
        book: 'libro',
        clock: 'reloj',
        vase: 'florero',
        scissors: 'tijeras',
        'teddy bear': 'oso de peluche',
        'hair drier': 'secador de pelo',
        toothbrush: 'cepillo de dientes'
    },
    fra_Latn: {
        person: 'personne',
        bicycle: 'velo',
        car: 'voiture',
        motorcycle: 'moto',
        airplane: 'avion',
        bus: 'bus',
        train: 'train',
        truck: 'camion',
        boat: 'bateau',
        'traffic light': 'feu de circulation',
        'stop sign': 'panneau stop',
        bench: 'banc',
        bird: 'oiseau',
        cat: 'chat',
        dog: 'chien',
        horse: 'cheval',
        sheep: 'mouton',
        cow: 'vache',
        backpack: 'sac a dos',
        umbrella: 'parapluie',
        handbag: 'sac a main',
        tie: 'cravate',
        suitcase: 'valise',
        'sports ball': 'ballon',
        bottle: 'bouteille',
        'wine glass': 'verre a vin',
        cup: 'tasse',
        fork: 'fourchette',
        knife: 'couteau',
        spoon: 'cuillere',
        bowl: 'bol',
        banana: 'banane',
        apple: 'pomme',
        sandwich: 'sandwich',
        orange: 'orange',
        broccoli: 'brocoli',
        carrot: 'carotte',
        pizza: 'pizza',
        donut: 'beignet',
        cake: 'gateau',
        chair: 'chaise',
        couch: 'canape',
        bed: 'lit',
        table: 'table',
        toilet: 'toilettes',
        tv: 'television',
        laptop: 'ordinateur portable',
        mouse: 'souris',
        remote: 'telecommande',
        keyboard: 'clavier',
        'cell phone': 'telephone portable',
        book: 'livre',
        clock: 'horloge',
        vase: 'vase',
        scissors: 'ciseaux',
        toothbrush: 'brosse a dents'
    },
    zho_Hans: {
        person: 'ren',
        car: 'qiche',
        bus: 'gonggongqiche',
        train: 'huoche',
        truck: 'ka che',
        bird: 'niao',
        cat: 'mao',
        dog: 'gou',
        horse: 'ma',
        backpack: 'beibao',
        umbrella: 'yu san',
        bottle: 'pingzi',
        cup: 'beizi',
        fork: 'chazi',
        knife: 'dao',
        spoon: 'shaozi',
        bowl: 'wan',
        banana: 'xiangjiao',
        apple: 'pingguo',
        orange: 'chengzi',
        carrot: 'huluobo',
        pizza: 'pisa',
        cake: 'dangao',
        chair: 'yizi',
        couch: 'shafa',
        bed: 'chuang',
        table: 'zhuozi',
        tv: 'dianshi',
        laptop: 'bijibendiannao',
        mouse: 'shubiao',
        keyboard: 'jianpan',
        'cell phone': 'shouji',
        book: 'shu',
        clock: 'zhong',
        toothbrush: 'yashua'
    },
    hin_Deva: {
        person: 'vyakti',
        bicycle: 'saikil',
        car: 'kar',
        bus: 'bas',
        train: 'relgadi',
        truck: 'trak',
        bird: 'pakshi',
        cat: 'billi',
        dog: 'kutta',
        horse: 'ghoda',
        backpack: 'basta',
        umbrella: 'chhata',
        bottle: 'botal',
        cup: 'cup',
        fork: 'kanta',
        knife: 'chaku',
        spoon: 'chammach',
        bowl: 'katora',
        banana: 'kela',
        apple: 'seb',
        orange: 'santra',
        carrot: 'gajar',
        pizza: 'pizza',
        cake: 'cake',
        chair: 'kursi',
        couch: 'sofa',
        bed: 'bistar',
        table: 'mez',
        tv: 'TV',
        laptop: 'laptop',
        mouse: 'mouse',
        keyboard: 'keyboard',
        'cell phone': 'mobile phone',
        book: 'kitaab',
        clock: 'ghadi',
        toothbrush: 'toothbrush'
    },
    afr_Latn: {
        person: 'persoon',
        bicycle: 'fiets',
        car: 'motor',
        bus: 'bus',
        train: 'trein',
        truck: 'vragmotor',
        bird: 'voel',
        cat: 'kat',
        dog: 'hond',
        horse: 'perd',
        backpack: 'rugsak',
        umbrella: 'sambreel',
        bottle: 'bottel',
        cup: 'koppie',
        fork: 'vurk',
        knife: 'mes',
        spoon: 'lepel',
        bowl: 'bak',
        banana: 'piesang',
        apple: 'appel',
        orange: 'lemoen',
        carrot: 'wortel',
        pizza: 'pizza',
        cake: 'koek',
        chair: 'stoel',
        couch: 'bank',
        bed: 'bed',
        table: 'tafel',
        tv: 'televisie',
        laptop: 'skootrekenaar',
        mouse: 'muis',
        keyboard: 'sleutelbord',
        'cell phone': 'selfoon',
        book: 'boek',
        clock: 'horlosie',
        toothbrush: 'tandeborsel'
    },
    zul_Latn: {
        person: 'umuntu',
        bicycle: 'ibhayisikili',
        car: 'imoto',
        bus: 'ibhasi',
        train: 'isitimela',
        truck: 'iloli',
        bird: 'inyoni',
        cat: 'ikati',
        dog: 'inja',
        horse: 'ihhashi',
        backpack: 'ubhaka',
        umbrella: 'isambulela',
        bottle: 'ibhodlela',
        cup: 'inkomishi',
        fork: 'imfoloko',
        knife: 'ummese',
        spoon: 'isipuni',
        bowl: 'isitsha',
        banana: 'ubhanana',
        apple: 'i-apula',
        orange: 'iwolintshi',
        carrot: 'isanqante',
        pizza: 'iphiza',
        cake: 'ikhekhe',
        chair: 'isitulo',
        couch: 'usofa',
        bed: 'umbhede',
        table: 'itafula',
        tv: 'ithelevishini',
        laptop: 'ikhompyutha ephathwayo',
        mouse: 'igundane',
        keyboard: 'ikhibhodi',
        'cell phone': 'umakhalekhukhwini',
        book: 'incwadi',
        clock: 'iwashi',
        toothbrush: 'isixubho'
    }
};

function translateLabel(label) {
    const normalizedLabel = label.toLowerCase().replaceAll('-', ' ');
    return labelTranslations[targetLanguage]?.[normalizedLabel] ?? label;
}

function getBoxArea(result) {
    const box = result.box;
    const xmin = Array.isArray(box) ? box[0] : box?.xmin;
    const ymin = Array.isArray(box) ? box[1] : box?.ymin;
    const xmax = Array.isArray(box) ? box[2] : box?.xmax;
    const ymax = Array.isArray(box) ? box[3] : box?.ymax;

    if (![xmin, ymin, xmax, ymax].every(Number.isFinite)) {
        return 0;
    }

    return Math.max(0, xmax - xmin) * Math.max(0, ymax - ymin);
}

// Helper to send progress updates to UI
const progressCallback = (progress) => {
    if (progress.status === 'progress') {
        postMessage({ 
            type: 'status', 
            data: `Downloading ${progress.file}: ${Math.round(progress.progress)}%` 
        });
    } else if (progress.status === 'done') {
        postMessage({ type: 'status', data: `Loaded ${progress.file}` });
    }
};

async function initModels() {
    try {
        postMessage({ type: 'status', data: 'Initializing Vision Model (DETR)...' });
        objectDetector = await pipeline('object-detection', 'Xenova/detr-resnet-50', {
            device: 'webgpu',
            progress_callback: progressCallback
        });

        postMessage({ type: 'status', data: 'ready' });
    } catch (err) {
        console.error("Model initialization failed:", err);
        postMessage({ type: 'status', data: 'Error loading models: ' + err.message });
    }
}

initModels();

self.addEventListener('message', async (event) => {
    const { type, image, language, threshold } = event.data;

    if (type === 'set-language') {
        if (targetLanguage !== language) {
            targetLanguage = language;
        }
    }

    if (type === 'set-threshold') {
        detectionThreshold = threshold;
    }

    if (type === 'set-focus-mode') {
        focusMode = event.data.enabled;
    }

    if (type === 'detect' && objectDetector) {
        let imageUrl = null;

        try {
            imageUrl = URL.createObjectURL(image);
            const detectionResults = await objectDetector(imageUrl, {
                threshold: detectionThreshold,
                percentage: false 
            });

            const translatedResults = detectionResults
                .map((res) => ({
                    ...res,
                    area: getBoxArea(res),
                    translated_label: translateLabel(res.label)
                }))
                .filter((res) => res.area >= 2500)
                .filter((res) => !focusMode || res.label !== 'person')
                .sort((a, b) => (b.score * b.area) - (a.score * a.area))
                .slice(0, focusMode ? 1 : 3);

            postMessage({ type: 'results', data: translatedResults });
        } catch (err) {
            console.error("Inference error:", err);
            postMessage({ type: 'results', data: [] });
        } finally {
            if (imageUrl) {
                URL.revokeObjectURL(imageUrl);
            }
        }
    }
});