Upload folder using huggingface_hub

src/orchestrator.py

@@ -206,7 +206,8 @@ def run_inspection(pil_img: Image.Image,
 
     # ── STEP 7: Index 2 retrieval ─────────────────────────────
     similar_cases = retriever.retrieve_similar_defects(
-        clip_crop, k=5, exclude_hash=image_hash
+        clip_crop, k=5, exclude_hash=image_hash,
+        category_filter=category
     )
 
     # ── STEP 8: Knowledge graph traversal ────────────────────

src/patchcore.py

@@ -160,17 +160,16 @@ class PatchCoreExtractor:
 
     def get_anomaly_centroid(self, heatmap: np.ndarray) -> tuple:
         """
-        Find centroid of highest-activation region.
+        Find the peak (highest activation) location of the anomaly.
         Used to locate defect crop for Index 2 retrieval.
-        Returns: (cx, cy) pixel coordinates
+        Returns: (cx, cy) pixel coordinates of maximum activation
         """
-        threshold = np.percentile(heatmap, 90)
-        mask = heatmap > threshold
-        if mask.sum() == 0:
+        if heatmap.size == 0:
             return (112, 112)   # centre fallback
-
-        ys, xs = np.where(mask)
-        return (int(xs.mean()), int(ys.mean()))
+        
+        # Use peak location, not mean of thresholded region
+        max_idx = np.unravel_index(np.argmax(heatmap), heatmap.shape)
+        return (int(max_idx[1]), int(max_idx[0]))  # cx, cy
 
     def calibrate_score(self,
                          raw_score: float,

src/retriever.py

@@ -99,11 +99,13 @@ class FAISSRetriever:
     def retrieve_similar_defects(self,
                                   clip_crop_embedding: np.ndarray,
                                   k: int = 5,
-                                  exclude_hash: str = None) -> list:
+                                  exclude_hash: str = None,
+                                  category_filter: str = None) -> list:
         """
         Given a defect-crop CLIP embedding, return k most similar
         historical defect cases.
         exclude_hash: skip self-match (same image submitted again)
+        category_filter: only return cases from specified category
         Returns: list of metadata dicts with similarity scores
         """
         query = clip_crop_embedding.reshape(1, -1).astype(np.float32)
@@ -118,6 +120,9 @@ class FAISSRetriever:
                 continue
             meta = self.index2_metadata[idx].copy()
             meta["similarity_score"] = float(dist)
+            # Filter by category if provided
+            if category_filter and meta.get("category") != category_filter:
+                continue
             # Skip self-match
             if exclude_hash and meta.get("image_hash") == exclude_hash:
                 continue