MediaWiki:Gadget-LabelScan.js: Unterschied zwischen den Versionen

Version vom 9. November 2025, 22:38 Uhr

/* global mw */
(() => {
  'use strict';

  const CFG = {
    // ---- Daten & Model ----
    indexTitle: (window.LabelScanConfig && window.LabelScanConfig.indexTitle) ||
                'MediaWiki:Gadget-LabelScan-index.json',
    transformersURL: 'https://cdn.jsdelivr.net/npm/@xenova/transformers@2.15.0',
    modelId: 'Xenova/clip-vit-base-patch32',
    localModelPath: '/models',          // <<— deine Modelle liegen hier
    topK: 8,                            // wie viele Treffer insgesamt
    maxSide: 1280,                      // Downscale vor Auto-Crop (Performance)

    // ---- Auto-Crop (Feature 4) ----
    autoCrop: true,
    edgeKeepRatio: 0.10,                // oberste 10% Kanten als Maske
    cropPadding: 0.08,                  // 8% Randzugabe um die Box
    cropMinRel: 0.40,                   // min. 40% der kleineren Bildkante

    // ---- Score-Badges (Feature 3) ----
    showNumericScore: false,            // true = Zahlen zeigen, false = Badges
    confidenceBands: [0.90, 0.80],      // hoch ≥0.90, mittel ≥0.80, sonst niedrig

    // ---- Sonstiges ----
    debug: true
  };

  // --------- Helpers ----------
  const log=(...a)=>{ if(CFG.debug) console.log('[LabelScan]',...a); };
  const warn=(...a)=>{ if(CFG.debug) console.warn('[LabelScan]',...a); };
  const err=(...a)=>{ console.error('[LabelScan]',...a); };
  const qs=id=>document.getElementById(id);
  const setStatus=t=>{ const el=qs('ados-scan-status'); if(el) el.textContent=t||''; };
  const setProgress=p=>{
    const bar=qs('ados-scan-progress'); if(!bar) return;
    if(p==null){ bar.hidden=true; bar.value=0; }
    else{ bar.hidden=false; bar.value=Math.max(0,Math.min(1,p)); }
  };

  function showPreview(file){
    const url=URL.createObjectURL(file);
    const prev=qs('ados-scan-preview');
    if(prev){
      prev.innerHTML=`<img alt="Vorschau" style="max-width:100%;height:auto;border-radius:8px;" src="${url}">`;
      prev.setAttribute('aria-hidden','false');
    }
  }

  function base64ToFloat32(b64){
    const bin=atob(b64), len=bin.length;
    const buf=new ArrayBuffer(len), view=new Uint8Array(buf);
    for(let i=0;i<len;i++) view[i]=bin.charCodeAt(i);
    return new Float32Array(buf);
  }

  // --------- Index ----------
  let INDEX=[], INDEX_EMB=[];
  async function loadIndex({ ui=true }={}){
    if(INDEX.length) return INDEX;
    if(ui){ setStatus('Index laden …'); setProgress(0.03); }
    const rawURL = mw.util.getUrl(CFG.indexTitle,{ action:'raw', ctype:'application/json' });
    const res = await fetch(rawURL,{ cache:'reload' });
    if(!res.ok) throw new Error('Index nicht ladbar: '+res.status);
    const json = await res.json();
    if(!Array.isArray(json)) throw new Error('Index ist keine Array-JSON');
    INDEX = json;
    INDEX_EMB = INDEX.map(it => (typeof it.embed==='string' && it.embed.length) ? base64ToFloat32(it.embed) : null);
    log('Index geladen:', INDEX.length, 'Einträge');
    log('Embeddings vorhanden:', INDEX_EMB.filter(v=>v&&v.length).length, '/', INDEX.length);
    if(ui) setProgress(0.06);
    return INDEX;
  }

  // --------- Transformers (lokal) ----------
  let _visionLoadPromise=null;
  async function ensureClipVision(){
    if(_visionLoadPromise) return _visionLoadPromise;

    setStatus('Modell laden …'); setProgress(0.08);

    _visionLoadPromise = (async()=>{
      const mod = await import(/* webpackIgnore: true */ CFG.transformersURL);

      // Nur lokal laden
      mod.env.allowLocalModels = true;
      mod.env.allowRemoteModels = false;
      mod.env.localModelPath   = CFG.localModelPath;

      // (Optional) WebGPU bevorzugen – fallback bleibt wasm
      // mod.env.backends = mod.env.backends || {};
      // mod.env.backends.onnx = mod.env.backends.onnx || {};
      // mod.env.backends.onnx.preferredBackend = 'webgpu';

      // WASM-Runtime (ort-wasm-simd.wasm) von transformers-CDN
      mod.env.backends = mod.env.backends || {};
      mod.env.backends.onnx = mod.env.backends.onnx || {};
      mod.env.backends.onnx.wasm = mod.env.backends.onnx.wasm || {};
      mod.env.backends.onnx.wasm.wasmPaths =
        'https://cdn.jsdelivr.net/npm/@xenova/transformers@2.15.0/dist/';

      const [processor, model] = await Promise.all([
        mod.AutoProcessor.from_pretrained(CFG.modelId),
        mod.CLIPVisionModelWithProjection.from_pretrained(CFG.modelId, { quantized: true })
      ]);

      let backend='unknown';
      try { backend = model?.session?.executionProvider || backend; } catch(_){}
      log('CLIP ready (vision, local):', model?.constructor?.name || 'unknown', '| Backend:', backend);

      return { mod, processor, model };
    })();

    return _visionLoadPromise;
  }

  // --------- Auto-Crop Heuristik ----------
  function toCanvasScaled(img, maxSide){
    const c=document.createElement('canvas');
    let { width:w, height:h } = img;
    const s = Math.min(1, maxSide / Math.max(w,h));
    w = Math.round(w*s); h = Math.round(h*s);
    c.width=w; c.height=h;
    const g=c.getContext('2d', { willReadFrequently:true });
    g.imageSmoothingEnabled = true;
    g.drawImage(img,0,0,w,h);
    return c;
  }

  function autoCropCanvas(inCanvas){
    const w=inCanvas.width, h=inCanvas.height;
    const ctx=inCanvas.getContext('2d', { willReadFrequently:true });
    const imgData=ctx.getImageData(0,0,w,h);
    const data=imgData.data;

    // Graustufen
    const gray=new Uint8ClampedArray(w*h);
    for(let y=0, p=0, i=0; y<h; y++){
      for(let x=0; x<w; x++, i++, p+=4){
        const r=data[p], g=data[p+1], b=data[p+2];
        gray[i] = (0.299*r + 0.587*g + 0.114*b)|0;
      }
    }

    // Sobel-Kanten (Magnitude)
    const mag=new Float32Array(w*h);
    const sobelX=[-1,0,1,-2,0,2,-1,0,1];
    const sobelY=[-1,-2,-1,0,0,0,1,2,1];
    for(let y=1; y<h-1; y++){
      for(let x=1; x<w-1; x++){
        let gx=0, gy=0, k=0;
        for(let j=-1;j<=1;j++){
          for(let i=-1;i<=1;i++,k++){
            const v=gray[(y+j)*w + (x+i)];
            gx += sobelX[k]*v; gy += sobelY[k]*v;
          }
        }
        mag[y*w+x] = Math.hypot(gx,gy);
      }
    }

    // Schwellwert: oberes x%-Quantil
    const vals = Array.from(mag).sort((a,b)=>a-b);
    const keep = CFG.edgeKeepRatio;
    const tIdx = Math.max(0, Math.min(vals.length-1, Math.floor(vals.length*(1-keep))));
    const thr = vals[tIdx];

    // Bounding-Box der Pixel > thr
    let minX=w, minY=h, maxX=0, maxY=0, count=0;
    for(let y=0;y<h;y++){
      for(let x=0;x<w;x++){
        const m=mag[y*w+x];
        if(m>thr){ count++; if(x<minX)minX=x; if(y<minY)minY=y; if(x>maxX)maxX=x; if(y>maxY)maxY=y; }
      }
    }

    if(count<50) return inCanvas; // zu wenig Signal → return original

    // Padding
    const pad = Math.round(CFG.cropPadding * Math.max(w,h));
    minX = Math.max(0, minX - pad);
    minY = Math.max(0, minY - pad);
    maxX = Math.min(w-1, maxX + pad);
    maxY = Math.min(h-1, maxY + pad);

    // Mindestgröße
    const boxW=maxX-minX+1, boxH=maxY-minY+1;
    const minLen = Math.round(CFG.cropMinRel * Math.min(w,h));
    let cx=minX, cy=minY, cw=boxW, ch=boxH;
    if(cw<minLen || ch<minLen){
      const needW = Math.max(minLen, cw);
      const needH = Math.max(minLen, ch);
      const centerX = Math.round((minX+maxX)/2);
      const centerY = Math.round((minY+maxY)/2);
      cx = Math.max(0, Math.min(w-needW, centerX - Math.floor(needW/2)));
      cy = Math.max(0, Math.min(h-needH, centerY - Math.floor(needH/2)));
      cw = needW; ch = needH;
    }

    const out=document.createElement('canvas');
    out.width=cw; out.height=ch;
    out.getContext('2d').drawImage(inCanvas, cx, cy, cw, ch, 0, 0, cw, ch);
    return out;
  }

  // --------- Embedding-Pipeline ---------
  async function embedFileImage(file){
    const { mod, processor, model } = await ensureClipVision();

    setStatus('Bild vorbereiten …'); setProgress(0.20);

    // 1) Bild laden
    const img = await new Promise((res,rej)=>{
      const url=URL.createObjectURL(file);
      const image=new Image();
      image.crossOrigin='anonymous';
      image.onload=()=>{ URL.revokeObjectURL(url); res(image); };
      image.onerror=e=>{ URL.revokeObjectURL(url); rej(e); };
      image.src=url;
    });

    // 2) Scale → Auto-Crop
    let canvas = toCanvasScaled(img, CFG.maxSide);
    if(CFG.autoCrop){
      setStatus('Auto-Crop …'); setProgress(0.30);
      canvas = autoCropCanvas(canvas);
    }

    // 3) Canvas → Blob → RawImage (robust für Processor)
    const blob = await new Promise(r => canvas.toBlob(r, 'image/jpeg', 0.95));
    const imageRaw = await mod.RawImage.fromBlob(blob);

    setStatus('Bild analysieren …'); setProgress(0.45);

    // 4) Processor & Model
    const inputs = await processor(imageRaw, { return_tensors: 'pt' });
    const out = await model.forward({ pixel_values: inputs.pixel_values });

    const vec = out?.image_embeds?.data || out?.image_embeds;
    if(!(vec instanceof Float32Array)) throw new Error('Embedding-Format unerwartet');

    // 5) Normieren
    let n=0; for(let i=0;i<vec.length;i++) n+=vec[i]*vec[i];
    const norm = Math.sqrt(n)||1;
    const v = new Float32Array(vec.length);
    for(let i=0;i<vec.length;i++) v[i]=vec[i]/norm;
    return v;
  }

  function cosine(a,b){ let s=0,L=Math.min(a.length,b.length); for(let i=0;i<L;i++) s+=a[i]*b[i]; return s; }
  function rankByCosine(q){
    const s=[];
    for(let i=0;i<INDEX.length;i++){
      const v=INDEX_EMB[i];
      if(!v) continue;
      s.push({ i, score: cosine(q,v) });
    }
    s.sort((a,b)=> b.score-a.score);
    return s.slice(0, CFG.topK);
  }

  // --------- Score-Badges (Feature 3) ---------
  function scoreBadge(score){
    if (CFG.showNumericScore) {
      return `<span style="font-variant-numeric:tabular-nums;color:#666">${score.toFixed(3)}</span>`;
    }
    const [hi, mid] = CFG.confidenceBands || [0.90, 0.80];
    let txt = 'niedrig', bg = '#f1f5f9', fg = '#334155';
    if (score >= hi) { txt = 'hoch';   bg = '#e6ffed'; fg = '#0a7d2c'; }
    else if (score >= mid) { txt = 'mittel'; bg = '#fff7e6'; fg = '#a45500'; }
    return `<span style="display:inline-block;padding:.12rem .45rem;border-radius:999px;background:${bg};color:${fg};font-weight:600;font-size:.85em;line-height:1">${txt}</span>`;
  }

  // --------- Rendering (Top-3 Karten + Restliste) ---------
  function renderResults(ranked){
    const box=qs('ados-scan-results');
    if(!box) return;
    box.innerHTML='';

    if(!ranked || !ranked.length){
      box.innerHTML='<div class="empty">Keine klaren Treffer. Bitte ein anderes Foto oder näher am Frontlabel.</div>';
      return;
    }

    const makeCard = (it, score) => `
      <div class="ados-card" style="display:grid;grid-template-columns:120px 1fr;gap:14px;align-items:center;padding:12px;border:1px solid #e6e6e6;border-radius:14px;box-shadow:0 1px 8px rgba(0,0,0,.04);">
        ${it.thumb?`<img src="${it.thumb}" alt="" style="width:120px;height:auto;border-radius:10px;">`
                   :`<div style="width:120px;height:90px;background:#f3f3f3;border-radius:10px;"></div>`}
        <div style="display:flex;flex-direction:column;gap:8px;">
          <div style="font-weight:700;font-size:1.05rem;line-height:1.2;">
            <a href="${mw.util.getUrl((it.title||'').replace(/ /g,'_'))}">${mw.html.escape(it.title||'')}</a>
          </div>
          <div>${scoreBadge(score)}</div>
          <div>
            <a href="${mw.util.getUrl((it.title||'').replace(/ /g,'_'))}" class="mw-ui-button" style="display:inline-block;padding:.4rem .7rem;border-radius:8px;background:#2a4b8d;color:#fff;text-decoration:none;">Artikel öffnen</a>
          </div>
        </div>
      </div>`;

    const grid = document.createElement('div');
    grid.style.display='grid';
    grid.style.gridTemplateColumns='1fr';
    grid.style.gap='12px';

    // Top-3 als Karten
    ranked.slice(0,3).forEach(({i,score})=>{
      const it=INDEX[i];
      grid.innerHTML += makeCard(it, score);
    });

    // Rest als Liste
    const rest = ranked.slice(3);
    if(rest.length){
      const sep = document.createElement('div');
      sep.style.margin='12px 0 4px';
      sep.style.borderTop='1px solid #e8e8e8';
      sep.style.paddingTop='10px';
      sep.innerHTML = '<div style="font-weight:600;color:#444;">Weitere Treffer</div>';
      grid.appendChild(sep);

      rest.forEach(({i,score})=>{
        const it=INDEX[i];
        const link=mw.util.getUrl((it.title||'').replace(/ /g,'_'));
        const thumb=it.thumb||'';
        const row=document.createElement('div');
        row.className='ados-hit';
        row.style.display='grid';
        row.style.gridTemplateColumns='60px 1fr auto';
        row.style.alignItems='center';
        row.style.gap='10px';
        row.style.padding='.35rem 0';
        row.innerHTML=
          (thumb?`<img src="${thumb}" alt="" style="width:60px;height:auto;border-radius:6px;">`:'<div></div>')+
          `<div><b><a href="${link}">${mw.html.escape(it.title||'')}</a></b></div>`+
          `<div>${scoreBadge(score)}</div>`;
        grid.appendChild(row);
      });
    }

    box.appendChild(grid);
  }

  // --------- UI / Flow ----------
  let BOUND=false;
  function bindUI(){
    if(BOUND) return;

    const btnRun = qs('ados-scan-run');
    const inCam  = qs('ados-scan-file-camera');
    const inGal  = qs('ados-scan-file-gallery');
    const btnCam = qs('ados-scan-btn-camera');
    const btnGal = qs('ados-scan-btn-gallery');
    const drop   = qs('ados-scan-drop');
    const btnReset = qs('ados-scan-reset');

    if(!btnRun || !inCam || !inGal) return;

    btnCam && btnCam.addEventListener('click', ()=> inCam.click());
    btnGal && btnGal.addEventListener('click', ()=> inGal.click());
    const pick = e => { const f=e.target.files?.[0]; if(f) showPreview(f); };
    inCam.addEventListener('change', pick);
    inGal.addEventListener('change', pick);

    if(drop){
      drop.addEventListener('dragover', ev => { ev.preventDefault(); drop.classList.add('is-over'); });
      drop.addEventListener('dragleave', ()=> drop.classList.remove('is-over'));
      drop.addEventListener('drop', ev => {
        ev.preventDefault(); drop.classList.remove('is-over');
        const f = ev.dataTransfer?.files?.[0];
        if(f){
          const dt=new DataTransfer(); dt.items.add(f);
          inGal.files=dt.files; showPreview(f);
        }
      });
    }

    btnReset && btnReset.addEventListener('click', ()=>{
      setStatus('Bereit.'); setProgress(null);
      const p=qs('ados-scan-preview'); if(p) p.innerHTML='<div class="note">Noch keine Vorschau.</div>';
      const r=qs('ados-scan-results'); if(r) r.innerHTML='<div class="empty">Hier erscheinen Treffer.</div>';
      inCam.value=''; inGal.value='';
    });

    btnRun.addEventListener('click', onRunClick);

    BOUND=true; log('UI gebunden.');
  }

  async function onRunClick(){
    try{
      const btnRun = qs('ados-scan-run');
      const inCam  = qs('ados-scan-file-camera');
      const inGal  = qs('ados-scan-file-gallery');

      const file = inCam.files?.[0] || inGal.files?.[0];
      if(!file){ alert('Bitte zuerst ein Foto auswählen.'); return; }

      if(btnRun) btnRun.disabled = true;

      await loadIndex({ ui:true });
      await ensureClipVision(); // warmup

      const q = await embedFileImage(file);

      setProgress(0.70);
      setStatus('Abgleich …');

      const ranked = rankByCosine(q);
      renderResults(ranked);

      setStatus('Fertig.');
      setProgress(null);
    } catch(e){
      err('Fehler', e);
      setStatus('Fehler bei Erkennung.');
      setProgress(null);
    } finally {
      const btnRun = qs('ados-scan-run');
      if(btnRun) btnRun.disabled = false;
    }
  }

  function init(){
    if(document.readyState==='loading'){
      document.addEventListener('DOMContentLoaded', bindUI, { once: true });
    } else {
      bindUI();
    }
    loadIndex({ ui:false }).catch(err).finally(()=>{ setStatus('Bereit.'); setProgress(null); });
  }

  log('gadget file loaded');
  init();

})();