added detection boxes & bugfixes

backend/ml/predict_detector_model.py

@@ -23,7 +23,7 @@ def main():
     root = Path(args.root)
     image_path = Path(args.image)
 
-    model_path = root / "detector" / "best.pt"
+    model_path = root / "detector" / "model" / "best.pt"
     if not model_path.exists():
         print(json.dumps({
             "available": False,

backend/training.go

@@ -916,13 +916,16 @@ func trainingWriteDetectorSample(root string, sample *TrainingSample, boxes []Tr
 			continue
 		}
 
-		// Erwartung: box.X/Y/W/H sind bereits YOLO-normalisiert:
+		// Frontend/Predictor nutzen x/y als linke obere Ecke.
-		// x_center, y_center, width, height.
+		// YOLO erwartet x_center/y_center.
+		xCenter := clamp01(x + w/2)
+		yCenter := clamp01(y + h/2)
+
 		lines = append(lines, fmt.Sprintf(
 			"%d %.6f %.6f %.6f %.6f",
 			classID,
-			x,
+			xCenter,
-			y,
+			yCenter,
 			w,
 			h,
 		))

frontend/src/components/ui/TrainingTab.tsx

@@ -1,7 +1,7 @@
 // frontend/src/components/ui/TrainingTab.tsx
 'use client'
 
-import { useCallback, useEffect, useMemo, useState } from 'react'
+import { useCallback, useEffect, useMemo, useRef, useState } from 'react'
 import Button from './Button'
 import LoadingSpinner from './LoadingSpinner'
 import { formatBytes, formatDuration } from './formatters'
@@ -60,6 +60,7 @@ type CorrectionState = {
   bodyPartsPresent: string[]
   objectsPresent: string[]
   clothingPresent: string[]
+  boxes: TrainingBox[]
 }
 
 type TrainingBox = {
@@ -99,20 +100,6 @@ function scoredLabelsText(items?: ScoredLabel[] | null) {
     .join(', ')
 }
 
-function detectionBoxes(sample: TrainingSample | null) {
-  const boxes = sample?.prediction.boxes ?? []
-
-  return boxes.filter((box) => {
-    const label = String(box.label || '').trim()
-    if (!label) return false
-
-    const isBodyPart = sample?.prediction.bodyPartsPresent?.some((x) => x.label === label)
-    const isObject = sample?.prediction.objectsPresent?.some((x) => x.label === label)
-
-    return isBodyPart || isObject
-  })
-}
-
 function clampPercent(v: number) {
   if (!Number.isFinite(v)) return 0
   return Math.max(0, Math.min(100, v))
@@ -133,6 +120,40 @@ function totalPeopleFromGenderCounts(state: Pick<CorrectionState, 'maleCount' |
   return safeCount(state.maleCount) + safeCount(state.femaleCount)
 }
 
+function clamp01(v: number) {
+  if (!Number.isFinite(v)) return 0
+  return Math.max(0, Math.min(1, v))
+}
+
+function normalizeBox(box: TrainingBox): TrainingBox {
+  const x = clamp01(box.x)
+  const y = clamp01(box.y)
+  const w = clamp01(Math.min(box.w, 1 - x))
+  const h = clamp01(Math.min(box.h, 1 - y))
+
+  return {
+    ...box,
+    x,
+    y,
+    w,
+    h,
+  }
+}
+
+function uniqStrings(values: string[]) {
+  const seen = new Set<string>()
+  const out: string[] = []
+
+  for (const value of values) {
+    const clean = String(value || '').trim()
+    if (!clean || seen.has(clean)) continue
+    seen.add(clean)
+    out.push(clean)
+  }
+
+  return out
+}
+
 function predictionToCorrection(sample: TrainingSample | null): CorrectionState {
   const p = sample?.prediction
 
@@ -148,6 +169,14 @@ function predictionToCorrection(sample: TrainingSample | null): CorrectionState
     bodyPartsPresent: (p?.bodyPartsPresent ?? []).map((x) => x.label),
     objectsPresent: (p?.objectsPresent ?? []).map((x) => x.label),
     clothingPresent: (p?.clothingPresent ?? []).map((x) => x.label),
+    boxes: (p?.boxes ?? []).map((box) => ({
+      label: String(box.label || '').trim(),
+      score: box.score,
+      x: clamp01(Number(box.x)),
+      y: clamp01(Number(box.y)),
+      w: clamp01(Number(box.w)),
+      h: clamp01(Number(box.h)),
+    })).filter((box) => box.label && box.w > 0 && box.h > 0),
   }
 }
 
@@ -184,6 +213,25 @@ export default function TrainingTab() {
   const [deletingTrainingData, setDeletingTrainingData] = useState(false)
   const [error, setError] = useState<string | null>(null)
   const [message, setMessage] = useState<string | null>(null)
+
+  const imageBoxRef = useRef<HTMLDivElement | null>(null)
+
+  const [drawingBox, setDrawingBox] = useState<TrainingBox | null>(null)
+  const [boxLabel, setBoxLabel] = useState('')
+
+  const boxLabels = useMemo(() => {
+    return uniqStrings([
+      ...labels.bodyParts,
+      ...labels.objects,
+      ...labels.clothing,
+    ]).sort((a, b) => a.localeCompare(b, undefined, { sensitivity: 'base' }))
+  }, [labels.bodyParts, labels.objects, labels.clothing])
+
+  const correctionBoxes = correction.boxes ?? []
+
+  const visibleBoxes = drawingBox
+    ? [...correctionBoxes, drawingBox]
+    : correctionBoxes
   
   const computedPeopleCount = useMemo(
     () => totalPeopleFromGenderCounts(correction),
@@ -195,8 +243,6 @@ export default function TrainingTab() {
     return `${sample.frameUrl}&t=${encodeURIComponent(sample.sampleId)}`
   }, [sample])
 
-  const boxes = useMemo(() => detectionBoxes(sample), [sample])
-
   const canStartTraining = Boolean(trainingStatus?.canTrain)
   const feedbackCount = trainingStatus?.feedbackCount ?? 0
   const requiredCount = trainingStatus?.requiredCount ?? 5
@@ -248,6 +294,11 @@ export default function TrainingTab() {
     })
   }, [])
 
+  useEffect(() => {
+    if (boxLabel && boxLabels.includes(boxLabel)) return
+    setBoxLabel(boxLabels[0] || '')
+  }, [boxLabel, boxLabels])
+
   useEffect(() => {
     void loadLabels()
     void loadNext()
@@ -267,12 +318,17 @@ export default function TrainingTab() {
           ...correction,
           peopleCount: totalPeopleFromGenderCounts(correction),
           unknownCount: 0,
+          boxes: (correction.boxes ?? [])
+            .map(normalizeBox)
+            .filter((box) => box.label && box.w > 0 && box.h > 0),
         }
 
         const payload = {
           sampleId: sample.sampleId,
           accepted,
-          correction: accepted ? undefined : correctionPayload,
+          correction: accepted && correctionPayload.boxes.length === 0
+            ? undefined
+            : correctionPayload,
         }
 
         const res = await fetch('/api/training/feedback', {
@@ -363,6 +419,86 @@ export default function TrainingTab() {
     }
   }, [loadNext, loadTrainingStatus, requiredCount])
 
+  const getPointerPosInImage = useCallback((clientX: number, clientY: number) => {
+    const el = imageBoxRef.current
+    if (!el) return null
+
+    const rect = el.getBoundingClientRect()
+    if (rect.width <= 0 || rect.height <= 0) return null
+
+    return {
+      x: clamp01((clientX - rect.left) / rect.width),
+      y: clamp01((clientY - rect.top) / rect.height),
+    }
+  }, [])
+
+  const startDrawBox = useCallback((e: React.PointerEvent<HTMLDivElement>) => {
+    if (!boxLabel) return
+    if (loading || saving) return
+
+    const pos = getPointerPosInImage(e.clientX, e.clientY)
+    if (!pos) return
+
+    e.currentTarget.setPointerCapture(e.pointerId)
+
+    setDrawingBox({
+      label: boxLabel,
+      x: pos.x,
+      y: pos.y,
+      w: 0,
+      h: 0,
+    })
+  }, [boxLabel, getPointerPosInImage, loading, saving])
+
+  const moveDrawBox = useCallback((e: React.PointerEvent<HTMLDivElement>) => {
+    if (!drawingBox) return
+
+    const pos = getPointerPosInImage(e.clientX, e.clientY)
+    if (!pos) return
+
+    const x1 = drawingBox.x
+    const y1 = drawingBox.y
+    const x2 = pos.x
+    const y2 = pos.y
+
+    setDrawingBox({
+      label: drawingBox.label,
+      x: Math.min(x1, x2),
+      y: Math.min(y1, y2),
+      w: Math.abs(x2 - x1),
+      h: Math.abs(y2 - y1),
+    })
+  }, [drawingBox, getPointerPosInImage])
+
+  const finishDrawBox = useCallback(() => {
+    if (!drawingBox) return
+
+    const box = normalizeBox(drawingBox)
+
+    setDrawingBox(null)
+
+    if (box.w < 0.01 || box.h < 0.01) return
+
+    setCorrection((prev) => ({
+      ...prev,
+      boxes: [...(prev.boxes ?? []), box],
+    }))
+  }, [drawingBox])
+
+  const removeBox = useCallback((index: number) => {
+    setCorrection((prev) => ({
+      ...prev,
+      boxes: (prev.boxes ?? []).filter((_, i) => i !== index),
+    }))
+  }, [])
+
+  const clearBoxes = useCallback(() => {
+    setCorrection((prev) => ({
+      ...prev,
+      boxes: [],
+    }))
+  }, [])
+
   return (
     <div className="grid grid-cols-1 items-stretch gap-3 lg:grid-cols-[300px_minmax(0,1fr)_300px] xl:grid-cols-[320px_minmax(0,1fr)_320px]">
       {/* Sidebar links */}
@@ -372,6 +508,10 @@ export default function TrainingTab() {
             Training
           </div>
 
+          <div className="rounded-full bg-gray-100 px-2 py-0.5 text-[11px] font-medium text-gray-700 ring-1 ring-gray-200 dark:bg-white/10 dark:text-gray-200 dark:ring-white/10">
+            {feedbackCount}
+          </div>
+
           <Button
             size="sm"
             variant="primary"
@@ -389,18 +529,6 @@ export default function TrainingTab() {
           Bestätige oder korrigiere die Analyse. Jede Antwort wird als Trainingsdatenpunkt gespeichert.
         </div>
 
-        <div className="mt-4 rounded-lg bg-gray-50 p-2 ring-1 ring-black/5 dark:bg-white/5 dark:ring-white/10">
-          <div className="flex items-center justify-between gap-2">
-            <div className="text-xs font-medium text-gray-900 dark:text-white">
-              Feedback
-            </div>
-
-            <div className="rounded-full bg-gray-100 px-2 py-0.5 text-[11px] font-medium text-gray-700 ring-1 ring-gray-200 dark:bg-white/10 dark:text-gray-200 dark:ring-white/10">
-              {feedbackCount}
-            </div>
-          </div>
-        </div>
-
         <div className="mt-4 space-y-2 text-xs">
           <div className="rounded-lg bg-gray-50 p-2 ring-1 ring-black/5 dark:bg-white/5 dark:ring-white/10">
             <div className="font-medium text-gray-900 dark:text-white">Datei</div>
@@ -431,17 +559,9 @@ export default function TrainingTab() {
             Aktuelle Modell-Erkennung
           </div>
 
-          <div className="mt-1 text-[11px] text-gray-500 dark:text-gray-400">
-            {sample?.prediction.modelAvailable
-              ? `Modell aktiv: ${sample.prediction.source || 'lokal'}`
-              : 'Noch kein trainiertes Modell vorhanden.'}
-          </div>
-
           <div className="mt-2 space-y-1 text-xs text-gray-600 dark:text-gray-300">
-            <div>Personen: {sample?.prediction.peopleCount ?? '—'}</div>
             <div>Männlich: {sample?.prediction.maleCount ?? '—'}</div>
             <div>Weiblich: {sample?.prediction.femaleCount ?? '—'}</div>
-            <div>Unklar: {sample?.prediction.unknownCount ?? '—'}</div>
 
             <div>
               Position: {sample?.prediction.sexPosition ?? '—'}{' '}
@@ -460,6 +580,79 @@ export default function TrainingTab() {
               Kleidung: {scoredLabelsText(sample?.prediction.clothingPresent)}
             </div>
           </div>
+
+          <div className="mt-2 rounded-lg bg-gray-50 p-2 ring-1 ring-black/5 dark:bg-white/5 dark:ring-white/10">
+            <div className="flex items-center justify-between gap-2">
+              <div>
+                <div className="text-xs font-medium text-gray-700 dark:text-gray-200">
+                  Detector-Boxen
+                </div>
+                <div className="text-[10px] text-gray-500 dark:text-gray-400">
+                  Label wählen, dann im Bild ziehen.
+                </div>
+              </div>
+
+              <div className="rounded-full bg-white px-2 py-0.5 text-[11px] font-semibold text-gray-800 ring-1 ring-gray-200 dark:bg-gray-950 dark:text-gray-100 dark:ring-white/10">
+                {correctionBoxes.length}
+              </div>
+            </div>
+
+            <select
+              value={boxLabel}
+              onChange={(e) => setBoxLabel(e.target.value)}
+              disabled={boxLabels.length === 0}
+              className="mt-2 h-9 w-full rounded-md border border-gray-200 bg-white px-2 text-sm dark:border-white/10 dark:bg-gray-950"
+            >
+              {boxLabels.length === 0 ? (
+                <option value="">Keine Box-Labels</option>
+              ) : (
+                boxLabels.map((x) => (
+                  <option key={x} value={x}>
+                    {x}
+                  </option>
+                ))
+              )}
+            </select>
+
+            <div className="mt-2 space-y-1">
+              {correctionBoxes.length === 0 ? (
+                <div className="text-[11px] text-gray-500 dark:text-gray-400">
+                  Noch keine Boxen gezeichnet.
+                </div>
+              ) : (
+                correctionBoxes.map((box, index) => (
+                  <div
+                    key={`${box.label}-${index}`}
+                    className="flex items-center justify-between gap-2 rounded-md bg-white px-2 py-1 text-[11px] ring-1 ring-gray-200 dark:bg-gray-950 dark:ring-white/10"
+                  >
+                    <div className="min-w-0 truncate text-gray-700 dark:text-gray-200">
+                      {index + 1}. {box.label}
+                    </div>
+
+                    <button
+                      type="button"
+                      className="shrink-0 rounded px-1.5 py-0.5 text-red-600 hover:bg-red-50 dark:text-red-300 dark:hover:bg-red-500/10"
+                      onClick={() => removeBox(index)}
+                    >
+                      löschen
+                    </button>
+                  </div>
+                ))
+              )}
+            </div>
+
+            {correctionBoxes.length > 0 ? (
+              <Button
+                size="sm"
+                variant="secondary"
+                className="mt-2 w-full"
+                disabled={saving || loading}
+                onClick={clearBoxes}
+              >
+                Alle Boxen entfernen
+              </Button>
+            ) : null}
+          </div>
         </div>
 
         <div className="mt-4 space-y-2">
@@ -527,39 +720,69 @@ export default function TrainingTab() {
             />
           ) : imageSrc ? (
             <div className="relative flex h-full w-full items-center justify-center">
-              <img
+              <div
-                src={imageSrc}
+                ref={imageBoxRef}
-                alt="Training Frame"
+                className="relative inline-block max-h-[72dvh] max-w-full select-none"
-                className="max-h-[72dvh] w-full object-contain"
+                onPointerDown={startDrawBox}
-              />
+                onPointerMove={moveDrawBox}
+                onPointerUp={finishDrawBox}
+                onPointerCancel={finishDrawBox}
+              >
+                <img
+                  src={imageSrc}
+                  alt="Training Frame"
+                  draggable={false}
+                  className="block max-h-[72dvh] max-w-full object-contain"
+                />
 
-              {boxes.length > 0 ? (
+                <div className="absolute inset-0">
-                <div className="pointer-events-none absolute inset-0">
+                  {visibleBoxes.map((box, index) => {
-                  {boxes.map((box, index) => {
                     const left = clampPercent(box.x * 100)
                     const top = clampPercent(box.y * 100)
                     const width = clampPercent(box.w * 100)
                     const height = clampPercent(box.h * 100)
+                    const isDraft = drawingBox && index === visibleBoxes.length - 1
 
                     return (
-                      <div
+                      <button
-                        key={`${box.label}-${index}`}
+                        key={`${box.label}-${index}-${isDraft ? 'draft' : 'box'}`}
-                        className="absolute rounded border-2 border-emerald-400 shadow-[0_0_0_1px_rgba(0,0,0,0.75)]"
+                        type="button"
+                        className={[
+                          'absolute rounded border-2 bg-transparent hover:bg-transparent active:bg-transparent focus:bg-transparent shadow-[0_0_0_1px_rgba(0,0,0,0.75)]',
+                          isDraft
+                            ? 'pointer-events-none border-amber-400'
+                            : 'border-emerald-400 hover:border-red-400',
+                        ].join(' ')}
                         style={{
                           left: `${left}%`,
                           top: `${top}%`,
                           width: `${width}%`,
                           height: `${height}%`,
                         }}
+                        title={isDraft ? box.label : `${box.label} löschen`}
+                        onPointerDown={(e) => {
+                          e.preventDefault()
+                          e.stopPropagation()
+                        }}
+                        onClick={(e) => {
+                          e.preventDefault()
+                          e.stopPropagation()
+                          if (!isDraft) removeBox(index)
+                        }}
                       >
-                        <div className="absolute left-0 top-0 -translate-y-full rounded-t bg-emerald-400 px-1.5 py-0.5 text-[11px] font-semibold text-black shadow">
+                        <span
+                          className={[
+                            'absolute left-0 top-0 -translate-y-full rounded-t px-1.5 py-0.5 text-[11px] font-semibold text-black shadow',
+                            isDraft ? 'bg-amber-400' : 'bg-emerald-400',
+                          ].join(' ')}
+                        >
                           {box.label} {typeof box.score === 'number' ? percent(box.score) : ''}
-                        </div>
+                        </span>
-                      </div>
+                      </button>
                     )
                   })}
                 </div>
-              ) : null}
+              </div>
             </div>
           ) : (
             <div className="text-sm text-white/80">Kein Frame geladen</div>

frontend/src/components/ui/formatters.ts

@@ -72,4 +72,5 @@ export function formatDateTime(v?: string | number | Date | null): string {
     hour: '2-digit',
     minute: '2-digit',
   })
-}
+}
+