what is blessing my ears rn

pipeline/core.py

@@ -114,11 +114,8 @@ class SmokingDetectionPipeline:
                 self.current_future = self.executor.submit(self._run_inference, frame)
 
             # 3. Apply cached detections to current frame for visualization
-            # This ensures smooth video output even if inference runs at 10fps
-            frame.detections.extend(self.latest_detections)
-
-            # 4. Run NMS to clean up overlapping boxes before display
-            frame.detections = nms_by_class(frame.detections, iou_threshold=0.5)
+            # This ensures smooth video output even if inference runs at lower fps
+            frame.detections = self.latest_detections.copy()
 
             # Ticket dispatch: Only attach to ONE frame, then clear
             # This prevents duplicate alerts in sinks like FileAlertStore
@@ -146,6 +143,9 @@ class SmokingDetectionPipeline:
 
         # --- Step 1: Broad Phase (Person Detection) ---
         people = self.person_model.predict(frame.data)
+
+        people = nms_by_class(people, iou_threshold=0.7)
+
         detections.extend(people)
 
         if not people:
@@ -153,13 +153,13 @@ class SmokingDetectionPipeline:
             return False, detections, None
 
         # --- Step 2: Narrow Phase (Cigarette Candidate) ---
-        candidates = self._find_candidates(frame, people)
+        candidates = self._find_candidates(frame, people, detections)
 
         if not candidates:
             return True, detections, None
 
         # --- Step 3: Verification (Classifier) ---
-        best_hit = self._verify_candidates(frame, candidates)
+        best_hit = self._verify_candidates(frame, candidates, detections)
 
         ticket = None
         # --- Step 4: Action ---
@@ -188,6 +188,9 @@ class SmokingDetectionPipeline:
             )
             detections.append(verified_det)
 
+        # Clean up overlapping boxes from model output
+        detections = nms_by_class(detections, iou_threshold=0.5)
+
         dt = time.time() - t0
         if dt > 0.1:
             logger.debug("Inference took %.3fs (Slow)", dt)
@@ -195,9 +198,13 @@ class SmokingDetectionPipeline:
         return True, detections, ticket
 
     def _find_candidates(
-        self, frame: Frame, people: List[Detection]
+        self, frame: Frame, people: List[Detection], detections_out: List[Detection]
     ) -> List[Tuple[Detection, List[Detection], Tuple[int, int]]]:
-        """Runs cigarette detection on expanded person crops."""
+        """Runs cigarette detection on expanded person crops.
+
+        Args:
+            detections_out: List to append candidate detections to (thread-safe output)
+        """
         candidates = []
         h_img, w_img = frame.data.shape[:2]
         margin = 0.2  # 20% expansion
@@ -226,12 +233,13 @@ class SmokingDetectionPipeline:
             # 3. Detect
             cig_candidates = self.cigarette_model.predict(person_crop)
 
-            # 4. Filter Noise
+            # 4. Filter Noise + NMS (model can return overlapping boxes for same cig)
             cig_candidates = [
                 c
                 for c in cig_candidates
                 if (c.box.x2 - c.box.x1) >= 10 and (c.box.y2 - c.box.y1) >= 10
             ]
+            cig_candidates = nms_by_class(cig_candidates, iou_threshold=0.5)
 
             if cig_candidates:
                 candidates.append((person, cig_candidates, (ex1, ey1)))
@@ -250,14 +258,18 @@ class SmokingDetectionPipeline:
                         label="candidate_cig",
                         metadata=cig.metadata,
                     )
-                    frame.detections.append(global_cig)
+                    detections_out.append(global_cig)
 
         return candidates
 
     def _verify_candidates(
-        self, frame: Frame, candidates: List[Tuple]
+        self, frame: Frame, candidates: List[Tuple], detections_out: List[Detection]
     ) -> Optional[Tuple]:
-        """Runs verification on candidate crops."""
+        """Runs verification on candidate crops.
+
+        Args:
+            detections_out: List to append verified detections to (thread-safe output)
+        """
         confirmed_hits = []
 
         for person, cigs, (ex1, ey1) in candidates:
@@ -305,7 +317,7 @@ class SmokingDetectionPipeline:
                         label="verified_smoking",
                         metadata={"verifier_score": score},
                     )
-                    frame.detections.append(verified_det)
+                    detections_out.append(verified_det)
 
         if confirmed_hits:
             return max(confirmed_hits, key=lambda x: x[2])