@camera.ui/server 0.0.49 → 0.0.50-internal.0

package/src/camera/python/test/iou_test.py

@@ -0,0 +1,799 @@
+import time
+from typing import Any, TypedDict, cast
+
+from typing_extensions import NotRequired
+
+from ..iou import merge_detections
+
+BoundingBox = tuple[float, float, float, float]
+
+
+class Detection(TypedDict):
+    id: NotRequired[str]
+    label: str
+    confidence: float
+    boundingBox: BoundingBox
+    inputWidth: int
+    inputHeight: int
+    origWidth: int
+    origHeight: int
+
+
+def create_detection(
+    id: str,
+    label: str,
+    confidence: float,
+    bounding_box: BoundingBox,
+    input_width: int,
+    input_height: int,
+) -> Detection:
+    return {
+        "id": id,
+        "label": cast(Any, label),
+        "confidence": confidence,
+        "boundingBox": bounding_box,
+        "inputWidth": input_width,
+        "inputHeight": input_height,
+        "origWidth": input_width,
+        "origHeight": input_height,
+    }
+
+
+class TestDetection(TypedDict):
+    name: str
+    detections: list[Detection]
+
+
+# Fixed values for "Many small boxes" test
+many_small_boxes_data = [
+    (100, 100, 200, 200, 0.9),
+    (150, 150, 250, 250, 0.85),
+    (175, 175, 275, 275, 0.8),
+    (300, 300, 400, 400, 0.95),
+    (350, 350, 450, 450, 0.87),
+    (500, 100, 600, 200, 0.92),
+    (550, 150, 650, 250, 0.88),
+    (700, 700, 800, 800, 0.91),
+    (750, 750, 850, 850, 0.86),
+    (200, 500, 300, 600, 0.93),
+]
+
+cameraUiDetections: list[Detection] = [
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (490, 201, 505, 231),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (522, 196, 542, 218),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (140, 193, 159, 232),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (540, 183, 556, 206),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (504, 181, 525, 200),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+cameraUiDetections2: list[Detection] = [
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (351, 333, 372, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (382, 279, 398, 314),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (399, 276, 433, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (411, 95, 461, 275),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (401, 90, 419, 120),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (271, 73, 414, 324),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (232, 5, 260, 25),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+cameraUiDetections3: list[Detection] = [
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (319, 212, 354, 316),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (361, 176, 396, 267),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (324, 164, 367, 199),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (338, 102, 358, 124),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (287, 76, 405, 316),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+cameraUiDetections4: list[Detection] = [
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (402, 347, 436, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (49, 342, 81, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (126, 332, 148, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (348, 329, 386, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (409, 328, 436, 348),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (317, 302, 334, 326),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (347, 297, 358, 331),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (201, 294, 222, 311),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (275, 278, 299, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (312, 274, 333, 306),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (565, 268, 595, 299),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (200, 267, 218, 289),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (123, 232, 197, 316),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (297, 230, 327, 242),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (7, 217, 155, 343),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (250, 110, 345, 243),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (345, 70, 406, 326),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (376, 59, 396, 90),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (297, 58, 341, 124),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (550, 39, 606, 73),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (192, 27, 235, 80),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+cameraUiDetections5: list[Detection] = [
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (52, 324, 82, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (399, 307, 449, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (530, 294, 560, 311),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (156, 290, 227, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (314, 279, 334, 320),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (554, 219, 624, 300),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (0, 218, 227, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (204, 166, 297, 346),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (254, 106, 302, 189),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (260, 76, 285, 121),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (296, 74, 409, 360),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (302, 45, 365, 112),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (0, 43, 22, 71),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (631, 41, 640, 83),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (549, 37, 627, 75),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (439, 34, 478, 52),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (507, 33, 548, 55),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (374, 28, 418, 101),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "label": "motion",
+        "confidence": 1,
+        "boundingBox": (190, 23, 239, 84),
+        "inputWidth": 640,
+        "inputHeight": 360,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+cameraUiDetections6: list[Detection] = [
+    {
+        "id": "9e732521-9f9e-4840-a09b-b621981751fd",
+        "label": "person",
+        "confidence": 0.6787109375,
+        "boundingBox": (466.5625, 195.0, 639.0625, 639.375),
+        "inputWidth": 640,
+        "inputHeight": 640,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "id": "9644b70e-126e-4e17-9cdb-ef16f851a96e",
+        "label": "toothbrush",
+        "confidence": 0.5546875,
+        "boundingBox": (169.140625, 281.875, 271.484375, 515.0),
+        "inputWidth": 640,
+        "inputHeight": 640,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "id": "23e1aff7-c7f1-47f5-bc5d-affe8a852af7",
+        "label": "bottle",
+        "confidence": 0.4921875,
+        "boundingBox": (241.7529296875, 581.30859375, 254.4970703125, 624.31640625),
+        "inputWidth": 640,
+        "inputHeight": 640,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+    {
+        "id": "814448bb-14d8-4928-95c3-7b3427f971c4",
+        "label": "bottle",
+        "confidence": 0.3056640625,
+        "boundingBox": (77.578125, 521.328125, 140.703125, 639.296875),
+        "inputWidth": 640,
+        "inputHeight": 640,
+        "origWidth": 2304,
+        "origHeight": 1296,
+    },
+]
+
+extendedMixedLabelsTest: TestDetection = {
+    "name": "Extended Mixed Labels",
+    "detections": [
+        create_detection("1", "person", 0.9, (100, 100, 200, 200), 640, 360),
+        create_detection("2", "person", 0.85, (150, 150, 250, 250), 640, 360),
+        create_detection("3", "car", 0.8, (300, 300, 400, 400), 640, 360),
+        create_detection("4", "car", 0.75, (350, 350, 450, 450), 640, 360),
+        create_detection("5", "bicycle", 0.7, (500, 500, 550, 550), 640, 360),
+        create_detection("6", "motorcycle", 0.65, (520, 520, 570, 570), 640, 360),
+        create_detection("7", "", 0.6, (600, 600, 650, 650), 640, 360),
+        create_detection("8", "person", 0.95, (180, 180, 280, 280), 640, 360),
+    ],
+}
+
+touchingBoxesTest: TestDetection = {
+    "name": "Touching Boxes",
+    "detections": [
+        create_detection("1", "motion", 0.9, (0, 0, 100, 100), 640, 360),
+        create_detection("2", "motion", 0.8, (100, 0, 200, 100), 640, 360),
+        create_detection("3", "motion", 0.7, (0, 100, 100, 200), 640, 360),
+        create_detection("4", "motion", 0.6, (100, 100, 200, 200), 640, 360),
+        create_detection("5", "motion", 0.5, (200, 0, 300, 300), 640, 360),
+    ],
+}
+
+test_cases: list[TestDetection] = [
+    {
+        "name": "Original test case",
+        "detections": [
+            create_detection("1", "motion", 0.9, (0, 0, 1000, 1000), 1000, 1000),
+            create_detection("2", "motion", 0.8, (100, 100, 300, 300), 1000, 1000),
+            create_detection("3", "motion", 0.7, (400, 400, 600, 600), 1000, 1000),
+            create_detection("4", "motion", 0.6, (700, 700, 900, 900), 1000, 1000),
+            create_detection("5", "motion", 0.75, (800, 800, 1100, 1100), 1000, 1000),
+        ],
+    },
+    {
+        "name": "Multiple overlapping boxes",
+        "detections": [
+            create_detection("1", "motion", 0.9, (0, 0, 500, 500), 1000, 1000),
+            create_detection("2", "motion", 0.8, (250, 250, 750, 750), 1000, 1000),
+            create_detection("3", "motion", 0.7, (500, 500, 1000, 1000), 1000, 1000),
+        ],
+    },
+    {
+        "name": "Boxes around the edges",
+        "detections": [
+            create_detection("1", "motion", 0.9, (400, 400, 600, 600), 1000, 1000),
+            create_detection("2", "motion", 0.8, (0, 0, 200, 200), 1000, 1000),
+            create_detection("3", "motion", 0.7, (800, 0, 1000, 200), 1000, 1000),
+            create_detection("4", "motion", 0.6, (0, 800, 200, 1000), 1000, 1000),
+            create_detection("5", "motion", 0.5, (800, 800, 1000, 1000), 1000, 1000),
+        ],
+    },
+    {
+        "name": "Nested boxes",
+        "detections": [
+            create_detection("1", "motion", 0.9, (0, 0, 1000, 1000), 1000, 1000),
+            create_detection("2", "motion", 0.8, (200, 200, 800, 800), 1000, 1000),
+            create_detection("3", "motion", 0.7, (400, 400, 600, 600), 1000, 1000),
+        ],
+    },
+    {
+        "name": "Many small boxes",
+        "detections": [
+            create_detection(
+                str(i + 1),
+                "motion",
+                confidence,
+                (x1, y1, x2, y2),
+                1000,
+                1000,
+            )
+            for i, (x1, y1, x2, y2, confidence) in enumerate(many_small_boxes_data)
+        ],
+    },
+    {
+        "name": "Mixed Labels",
+        "detections": [
+            create_detection("1", "person", 0.9, (100, 100, 200, 200), 640, 360),
+            create_detection("2", "car", 0.8, (150, 150, 250, 250), 640, 360),
+            create_detection("3", "vehicle", 0.85, (180, 180, 280, 280), 640, 360),
+        ],
+    },
+    extendedMixedLabelsTest,
+    touchingBoxesTest,
+    {
+        "name": "camera.ui 1",
+        "detections": cameraUiDetections,
+    },
+    {
+        "name": "camera.ui 2",
+        "detections": cameraUiDetections2,
+    },
+    {
+        "name": "camera.ui 3",
+        "detections": cameraUiDetections3,
+    },
+    {
+        "name": "camera.ui 4",
+        "detections": cameraUiDetections4,
+    },
+    {
+        "name": "camera.ui 5",
+        "detections": cameraUiDetections5,
+    },
+    {
+        "name": "camera.ui 6",
+        "detections": cameraUiDetections6,
+    },
+]
+
+
+def calculate_avg_box_size(detections: list[Detection]) -> float:
+    total_size = sum(
+        (d["boundingBox"][2] - d["boundingBox"][0]) * (d["boundingBox"][3] - d["boundingBox"][1])
+        for d in detections
+    )
+    return total_size / len(detections) if detections else 0
+
+
+def run_tests(iou_threshold: float = 0.01, close_threshold: float = 160):
+    for test_case in test_cases:
+        print(f"\nRunning test case: {test_case['name']}")
+        print(f"Original Detections: {len(test_case['detections'])}")
+
+        start_time = time.time()
+        merged_detections = merge_detections(test_case["detections"], iou_threshold, close_threshold)
+        end_time = time.time()
+
+        print(f"Merged Detections: {len(merged_detections)}")
+        print(f"Execution time: {((end_time - start_time) * 1000):.3f} ms")
+
+        # Basic validation
+        if len(merged_detections) <= len(test_case["detections"]):
+            print("✅ Test passed: Number of merged detections is not greater than original")
+        else:
+            print("❌ Test failed: Number of merged detections is greater than original")
+
+        # Check if bounding boxes are valid
+        invalid_boxes = [
+            d
+            for d in merged_detections
+            if d["boundingBox"][0] > d["boundingBox"][2] or d["boundingBox"][1] > d["boundingBox"][3]
+        ]
+        if not invalid_boxes:
+            print("✅ Test passed: All merged bounding boxes are valid")
+        else:
+            print("❌ Test failed: Some merged bounding boxes are invalid")
+
+        # Calculate average bounding box size change
+        original_avg_size = calculate_avg_box_size(test_case["detections"])
+        merged_avg_size = calculate_avg_box_size(cast(list[Detection], merged_detections))
+        size_change_percent = (
+            ((merged_avg_size - original_avg_size) / original_avg_size * 100) if original_avg_size else 0
+        )
+        print(f"Average bounding box size change: {size_change_percent:.2f}%")
+
+        # Print bounding boxes for visual inspection
+        print("Merged Bounding Boxes:")
+        for i, d in enumerate(merged_detections, 1):
+            print(f"  {d['label']}: Box {i}: [{', '.join(map(str, d['boundingBox']))}]")
+
+        print("\n" + "=" * 50)
+
+
+if __name__ == "__main__":
+    run_tests()