Crop Images around traffic signs using polygon coordinates

Imagery, data, and integrations
1

Hi, I’m trying to crop images around traffic signs, I’m trying to do it with polygon coordinates but in the results there are always cropped pieces of roads or other random pieces of the image. Am I doing something wrong with the normalization? Or is this not possible?

This is my code:

    for el in glob.glob('D:/Dataset Custom Mapillary/images/*.jpg'):
        name = Path(el).stem

        with open('D:/Dataset Custom Mapillary/annotations_image/' + name + '.json', "r") as file:
            temp = geojson.load(file)
            image_id = temp[1][0]['id_image']

        image_url = f'https://graph.mapillary.com/{image_id}?access_token={codes.API_KEY}&fields=height,width,thumb_original_url'
        detections_url = f'https://graph.mapillary.com/{image_id}/detections?access_token={codes.API_KEY}&fields=geometry'

        # request the image: get the height, width, and a URL to the original resolution blurred image
        response = requests.get(image_url)
        image_data = response.json()
        height = image_data['height']
        width = image_data['width']
        jpeg_url = image_data['thumb_original_url']
        
        print(jpeg_url)

        # request the detection
        response = requests.get(detections_url)
        json = response.json()
        detection = json['data'][0]
        base64_string = detection['geometry']

        # decode from base64
        vector_data = base64.decodebytes(base64_string.encode('utf-8'))

        # decode the vector tile into detection geometry
        decoded_geometry = mapbox_vector_tile.decode(vector_data)

        # select just the coordinate xy pairs from this detection
        detection_coordinates = decoded_geometry['mpy-or']['features'][0]['geometry']['coordinates']
        print(detection_coordinates)

        # normalize by the 4096 extent, then multiply by image height and width to get true coordinate location
        pg = [[[int(x / 4096 * width), int(y / 4096 * height)] for x, y in tuple(coord_pair)] for coord_pair in detection_coordinates]
        print(pg)

        img = cv2.imread(el)
        pts = np.array(pg)

        ## (1) Crop the bounding rect
        rect = cv2.boundingRect(pts)
        x, y, w, h = rect
        cropped = img[y:y + h, x:x + w].copy()

        ## (2) make mask
        pts = pts - pts.min(axis=0)

        mask = np.zeros(cropped.shape[:2], np.uint8)
        cv2.drawContours(mask, [pts], -1, (255, 255, 255), -1, cv2.LINE_AA)

        ## (3) do bit-op
        dst = cv2.bitwise_and(cropped, cropped, mask=mask)

        ## (4) add the white background
        bg = np.ones_like(cropped, np.uint8) * 255
        cv2.bitwise_not(bg, bg, mask=mask)
        dst2 = bg + dst

        cv2.imwrite("D:/Dataset Custom Mapillary/test_crop/cropped_"+name+".jpg", cropped)
        cv2.imwrite("D:/Dataset Custom Mapillary/test_crop/mask_"+name+".jpg", mask)
        cv2.imwrite("D:/Dataset Custom Mapillary/test_crop/dst_"+name+".jpg", dst)
        cv2.imwrite("D:/Dataset Custom Mapillary/test_crop/dst2_"+name+".jpg", dst2)

This is the output:

4032 3024
[[[2934, 1632], [2997, 1632], [2997, 1520], [2934, 1520], [2934, 1632]]]
[[[2888, 1204], [2950, 1204], [2950, 1122], [2888, 1122], [2888, 1204]]]
1223074091455516

3264 2448
[[[333, 2596], [451, 2596], [451, 2399], [333, 2399], [333, 2596]]]
[[[265, 1551], [359, 1551], [359, 1433], [265, 1433], [265, 1551]]]
1381323478934088

2576 1932
[[[1665, 2620], [1759, 2620], [1759, 2492], [1665, 2492], [1665, 2620]]]
[[[1047, 1235], [1106, 1235], [1106, 1175], [1047, 1175], [1047, 1235]]]
169575281758353

4032 3024
[[[2757, 2729], [2957, 2729], [2957, 2473], [2757, 2473], [2757, 2729]]]
[[[2713, 2014], [2910, 2014], [2910, 1825], [2713, 1825], [2713, 2014]]]
181274350545769

4032 3024
[[[885, 2516], [942, 2516], [942, 2307], [885, 2307], [885, 2516]]]
[[[871, 1857], [927, 1857], [927, 1703], [871, 1703], [871, 1857]]]
248177517087493

4032 3024
[[[1390, 2296], [1540, 2296], [1540, 2074], [1390, 2074], [1390, 2296]]]
[[[1368, 1695], [1515, 1695], [1515, 1531], [1368, 1531], [1368, 1695]]]
278425803923946

3968 2976
[[[352, 3746], [424, 3746], [424, 3646], [352, 3646], [352, 3746]]]
[[[341, 2721], [410, 2721], [410, 2649], [341, 2649], [341, 2721]]]
296311645439373

3968 2976
[[[352, 3746], [424, 3746], [424, 3646], [352, 3646], [352, 3746]]]
[[[341, 2721], [410, 2721], [410, 2649], [341, 2649], [341, 2721]]]
296369235433614