为什么直播时根本检测不到物体？ [已关闭]

Question

我的目标是使用

drawBorderboxes()

通过摄像头直播（在 Pycharm 中）来检测对象。我的第 1 部分工作得很好。这是为了确保使用 pylon 进行视频直播。第 2 部分是我遇到的问题，关于通过传入图像对象而不是文件夹路径来调用模型。

我哪里出错了（虽然这段代码运行顺利）？

这是我当前的代码：

# Part 1: 

logger = logging.getLogger()
tl_factory = pylon.TlFactory.GetInstance()
devices = tl_factory.EnumerateDevices()

if len(devices) == 0:
    logger.info("No cameras found!")
    exit()

top_camera = pylon.InstantCamera(tl_factory.CreateDevice(devices[0]))
top_camera.Open()

converter = pylon.ImageFormatConverter()
converter.OutputPixelFormat = pylon.PixelType_BGR8packed
converter.OutputBitAlignment = pylon.OutputBitAlignment_MsbAligned

top_camera.OutputQueueSize = 2
top_camera.StartGrabbing(pylon.GrabStrategy_LatestImages)

try:
    counter = 0
    while (True):
        top_grabResult = top_camera.RetrieveResult(5000, pylon.TimeoutHandling_Return)
        while not top_grabResult.GrabSucceeded():
            logger.info("top grab failed, trying again")
            top_grabResult = top_camera.RetrieveResult(5000, pylon.TimeoutHandling_Return)

        if top_grabResult.GrabSucceeded():
            image = converter.Convert(top_grabResult)
            img = image.Array
            img = cv2.rotate(img, cv2.ROTATE_90_COUNTERCLOCKWISE)
            original_h = img.shape[0]
            original_w = img.shape[1]
            new_h = (1088 / 720) * (original_w)  # x/2160 = 1088/720 solve for x
            offset = int(original_h - new_h)
            img = img[offset:original_h, 0:original_w]
            top_img = cv2.resize(img, (720, 1088))
            # img = cv2.resize(img, (0, 0), fx=0.25, fy=0.25)
            # cv2.namedWindow('esc_to_quit', cv2.WINDOW_AUTOSIZE)
            # cv2.imshow('esc_to_quit', img)
            # cv2.waitKey()

        else:
            logger.info(f"Top camera failed to grab an image")
            top_img = np.zeros((1088, 720, 3))
            top_img = cv2.resize(top_img, (0, 0), fx=0.5, fy=0.5)

        concat_img = np.concatenate((top_img), axis=1)
        cv2.namedWindow('esc_to_quit', cv2.WINDOW_AUTOSIZE)
        cv2.imshow('esc_to_quit', top_img)
        cv2.waitKey(1)
        print(counter)
        counter += 1

finally:
    top_camera.Close()

####### Part 2: call the model by passing in an image object rather than a path to a folder ####

# Function to draw bounding boxes on the image
def drawBorderBoxes(classifiedItems, image):
    for classifiedItem in classifiedItems:
        xyxy = classifiedItem[0]
        label = classifiedItem[1]
        color = classifiedItem[2]
        plot_one_box(xyxy, image, label=label, color=color)

# Function for live stream object detection
def detect_objects_live_stream(model, cam_feed):
    obj_detect = ObjectDetection()
    obj_detect.setModelTypeAsYOLOv5()
    obj_detect.setModelPath(model)
    obj_detect.loadModel()

    while True:
        ret, img = cam_feed.read()

        if not ret:
            print("Camera feed not available. Exiting...")
            break

        # Convert the OpenCV image to PIL Image
        pil_image = Image.fromarray(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))

        # Perform object detection
        annotated_image, preds = obj_detect.detectObjectsFromImage(input_image=pil_image,
                                                                   input_type="array",
                                                                   output_type="array",
                                                                   display_percentage_probability=False,
                                                                   display_object_name=True)

        # Draw bounding boxes on the image
        drawBorderBoxes(preds, img)

        # Display the annotated image
        cv2.imshow("", img)

        if (cv2.waitKey(1) & 0xFF == ord("q")) or (cv2.waitKey(1) == 27):
            break

    cam_feed.release()
    cv2.destroyAllWindows()

if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--weights', nargs='+', type=str, default='weights/yolov5s.pt', help='model.pt path(s)')
    parser.add_argument('--source', type=str, default='0', help='source')  # file/folder, 0 for webcam
    parser.add_argument('--output', type=str, default='inference/output', help='output folder')  # output folder
    parser.add_argument('--img-size', type=int, default=640, help='inference size (pixels)')
    parser.add_argument('--conf-thres', type=float, default=0.5, help='object confidence threshold')
    parser.add_argument('--iou-thres', type=float, default=0.5, help='IOU threshold for NMS')
    parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--view-img', action='store_true', help='display results')
    parser.add_argument('--save-txt', action='store_true', help='save results to *.txt')
    parser.add_argument('--classes', nargs='+', type=int, help='filter by class: --class 0, or --class 0 2 3')
    parser.add_argument('--agnostic-nms', action='store_true', help='class-agnostic NMS')
    parser.add_argument('--augment', action='store_true', help='augmented inference')
    parser.add_argument('--update', action='store_true', help='update all models')
    opt = parser.parse_args()

    if not os.path.exists(opt.output):
        os.mkdir(opt.output)
    else:
        shutil.rmtree(opt.output)
        os.mkdir(opt.output)

    with torch.no_grad():
        if opt.update:  # update all models (to fix SourceChangeWarning)
            for opt.weights in ['yolov5s.pt', 'yolov5m.pt', 'yolov5l.pt', 'yolov5x.pt']:
                detect(opt=opt)
                strip_optimizer(opt.weights)
        else:
            detect(opt=opt)

    # Create the YOLOv5 model
    model = attempt_load(opt.weights, map_location=opt.device)
    model.half() if opt.device != 'cpu' else None

    # Start live stream object detection
    cam_feed = cv2.VideoCapture(0)  # Modify the source here for a different camera feed
    detect_objects_live_stream(model, cam_feed)