@camera.ui/camera-ui-opencv 0.0.17 → 0.0.19

package/package.json

@@ -1,7 +1,7 @@
 {
   "displayName": "OpenCV",
   "name": "@camera.ui/camera-ui-opencv",
-  "version": "0.0.17",
+  "version": "0.0.19",
   "description": "camera.ui opencv plugin",
   "author": "seydx (https://github.com/seydx/camera.ui)",
   "main": "./src/main.py",

package/requirements.txt

@@ -1,3 +1,3 @@
 numpy==1.26.4
 opencv-python-headless==4.10.0.84
-camera-ui-python-types==0.1.88
+camera-ui-python-types==0.1.91

package/src/main.py

@@ -27,7 +27,7 @@ from detector_defaults import (
     default_threshold,
     default_threshold_bs,
 )
-from opencv_utils import get_detections, get_detections_bs, get_detections_fd
+from opencv_utils import get_detections, get_detections_bs, get_detections_fd, merge_overlapping_boxes
 
 
 class OpenCV(MotionDetectionPlugin):
@@ -315,7 +315,9 @@ class OpenCV(MotionDetectionPlugin):
 
                 previous_frame = gray
 
-            for det in dets:
+            merged_boxes = merge_overlapping_boxes(dets)
+
+            for det in merged_boxes:
                 (x1, y1, x2, y2) = det
 
                 pt1: cv2.typing.Point = (int(x1), int(y1))

package/src/opencv_utils.py

@@ -9,6 +9,56 @@ DEFAULT_MASK_KERNEL = np.array((9, 9), dtype=np.uint8)
 DEFAULT_LEARNING_RATE = 0.08
 
 
+def merge_overlapping_boxes(
+    boxes: list[tuple[float, float, float, float]], overlap_threshold: float = 0.1
+) -> list[tuple[float, float, float, float]]:
+    if not boxes:
+        return []
+
+    boxes_array = np.array(boxes)
+    merged_boxes: list[tuple[float, float, float, float]] = []
+
+    while len(boxes_array) > 0:
+        box = boxes_array[0]
+        boxes_array: np.ndarray[Any, Any] = boxes_array[1:]
+
+        x1, y1, w1, h1 = box
+        x2 = x1 + w1
+        y2 = y1 + h1
+
+        indices_to_remove: list[int] = []
+        for i, (x1o, y1o, w1o, h1o) in enumerate(boxes_array):
+            x2o = x1o + w1o
+            y2o = y1o + h1o
+
+            xx1 = max(x1, x1o)
+            yy1 = max(y1, y1o)
+            xx2 = min(x2, x2o)
+            yy2 = min(y2, y2o)
+
+            w_overlap = max(0, xx2 - xx1)
+            h_overlap = max(0, yy2 - yy1)
+            overlap_area = w_overlap * h_overlap
+            area1 = w1 * h1
+            area2 = w1o * h1o
+
+            if area1 == 0 or area2 == 0:
+                continue
+            overlap_ratio = overlap_area / min(area1, area2)
+
+            if overlap_ratio > overlap_threshold:
+                x1 = min(x1, x1o)
+                y1 = min(y1, y1o)
+                x2 = max(x2, x2o)
+                y2 = max(y2, y2o)
+                indices_to_remove.append(i)
+
+        boxes_array = np.delete(boxes_array, indices_to_remove, axis=0)
+        merged_boxes.append((int(x1), int(y1), int(x2 - x1), int(y2 - y1)))
+
+    return merged_boxes
+
+
 def get_mask(
     frame1: np.ndarray[Any, Any],
     frame2: np.ndarray[Any, Any],