diff --git a/src/dashboard.py b/src/dashboard.py
index 152503a..0077d74 100644
--- a/src/dashboard.py
+++ b/src/dashboard.py
@@ -55,17 +55,16 @@ class DashboardOverlay:
         stabilize_progress: float = 0.0,
         waiting_duration: float = 0.0,
     ) -> np.ndarray:
-        out = frame.copy()
         if roi is not None:
-            self._draw_roi(out, roi)
-            self._draw_counting_zone(out, roi)
-        self._draw_detections(out, detections)
-        self._draw_stats(out, loading_count, unloading_count, batch_id, history)
+            self._draw_roi(frame, roi)
+            self._draw_counting_zone(frame, roi)
+        self._draw_detections(frame, detections)
+        self._draw_stats(frame, loading_count, unloading_count, batch_id, history)
         self._draw_system_state(
-            out, system_state, batch_duration, idle_timer, stabilize_progress,
+            frame, system_state, batch_duration, idle_timer, stabilize_progress,
             waiting_duration,
         )
-        return out
+        return frame
 
     def _draw_roi(self, frame: np.ndarray, roi: TruckROI) -> None:
         cv2.rectangle(
@@ -84,13 +83,15 @@ class DashboardOverlay:
         self, frame: np.ndarray, roi: TruckROI, margin: int = 20,
     ) -> None:
         y = roi.line_y
-        # Draw zone band (semi-transparent)
-        overlay = frame.copy()
-        cv2.rectangle(
-            overlay, (roi.x1, y - margin), (roi.x2, y + margin),
-            MAGENTA, -1,
-        )
-        cv2.addWeighted(overlay, 0.15, frame, 0.85, 0, frame)
+        h, w = frame.shape[:2]
+        # Blend only the narrow strip region, not the full frame
+        sy = max(0, y - margin)
+        ey = min(h, y + margin)
+        sx = max(0, roi.x1)
+        ex = min(w, roi.x2)
+        strip = frame[sy:ey, sx:ex].copy()
+        cv2.rectangle(strip, (0, 0), (ex - sx, ey - sy), MAGENTA, -1)
+        cv2.addWeighted(strip, 0.15, frame[sy:ey, sx:ex], 0.85, 0, frame[sy:ey, sx:ex])
         # Draw center line
         cv2.line(frame, (roi.x1, y), (roi.x2, y), MAGENTA, 2)
         cv2.putText(
@@ -188,12 +189,12 @@ class DashboardOverlay:
         # Get display info for current state
         label, color = STATE_DISPLAY.get(system_state, ("UNKNOWN", GRAY))
 
-        # Draw state bar background
+        # Draw state bar — blend only the bottom bar strip, not full frame
         bar_h = 36
         bar_y = h - bar_h
-        overlay = frame.copy()
-        cv2.rectangle(overlay, (0, bar_y), (w, h), (0, 0, 0), -1)
-        cv2.addWeighted(overlay, 0.7, frame, 0.3, 0, frame)
+        strip = frame[bar_y:h, 0:w].copy()
+        cv2.rectangle(strip, (0, 0), (w, bar_h), (0, 0, 0), -1)
+        cv2.addWeighted(strip, 0.7, frame[bar_y:h, 0:w], 0.3, 0, frame[bar_y:h, 0:w])
 
         # Draw colored indicator dot
         cv2.circle(frame, (20, bar_y + bar_h // 2), 8, color, -1)
diff --git a/src/pipeline.py b/src/pipeline.py
index 1d05501..ccd3b54 100644
--- a/src/pipeline.py
+++ b/src/pipeline.py
@@ -404,8 +404,12 @@ def run_merged_pipeline(
     if truck_model_config is not None:
         truck_shared = YOLO(truck_model_config.path)
         truck_det = BaseDetector(truck_shared, conf=truck_conf, class_filter=("truck",))
-    elif len(model_configs) == 1 and "truck" in (model_configs[0].known_classes or []):
-        truck_det = detectors[0]
+    # ponytail: first-match truck detector; could prefer dedicated truck model over sack model w/ truck class
+    else:
+        for i, cfg in enumerate(model_configs):
+            if "truck" in (cfg.known_classes or []):
+                truck_det = detectors[i]
+                break
 
     # Single tracker using first model's weights
     tracker_model = YOLO(model_configs[0].path)
@@ -426,6 +430,7 @@ def run_merged_pipeline(
     frame_idx = 0
     completed_batches = 0
     _frame_times: list[float] = []
+    sup_det_interval = 5  # run supplementary models every N frames
 
     def on_batch_end(record):
         nonlocal completed_batches
@@ -472,16 +477,23 @@ def run_merged_pipeline(
                     truck_centroid=truck_centroid,
                 )
 
-            # Run all detectors in parallel
+            # Detection + tracking
             all_detections: list[Detection] = []
             if batch_mgr.is_active:
-                with ThreadPoolExecutor(max_workers=len(detectors)) as pool:
-                    futures = {pool.submit(d.detect, frame): d for d in detectors}
-                    for future in as_completed(futures):
-                        try:
-                            all_detections.extend(future.result())
-                        except Exception:
-                            pass
+                # Primary: tracker gives detections WITH track_id for counting
+                primary = tracker.update(frame, [])
+                all_detections.extend(primary)
+
+                # Supplementary: run extra models every N frames for richer annotations
+                if len(detectors) > 1 and frame_idx % sup_det_interval == 0:
+                    with ThreadPoolExecutor(max_workers=len(detectors) - 1) as pool:
+                        extras = [d for i, d in enumerate(detectors) if i != 0]
+                        futures = {pool.submit(d.detect, frame): d for d in extras}
+                        for future in as_completed(futures):
+                            try:
+                                all_detections.extend(future.result())
+                            except Exception:
+                                pass
 
                 # Deduplicate
                 merged = _deduplicate_detections(all_detections)
@@ -493,7 +505,8 @@ def run_merged_pipeline(
                 else:
                     tracked = stable
 
-                counter.update(tracked)
+                # Count using only detections with track_id (tracker results)
+                counter.update([d for d in tracked if d.track_id is not None])
             else:
                 tracked = []