Error with YOLOv10 video inference

[tuannhtn]

I could run your code with TensorRT 10 on images, but when running on video file I got this error:
Traceback (most recent call last):
File "D:\study\yolov10\TensorRT-For-YOLO-Series-cuda-python\trt.py", line 35, in
pred.detect_video(video, conf=0.1, end2end=args.end2end) # set 0 use a webcam
File "D:\study\yolov10\TensorRT-For-YOLO-Series-cuda-python\utils\utils.py", line 118, in detect_video
dets = np.concatenate([np.array(final_boxes)[:int(num[0])], np.array(final_scores)[:int(num[0])], np.array(final_cls_inds)[:int(num[0])]], axis=-1)
File "<array_function internals>", line 200, in concatenate
ValueError: all the input array dimensions except for the concatenation axis must match exactly, but along dimension 0, the array at index 0 has size 300 and the array at index 1 has size 1
Could anyone encountered this error and knew how to fix it? Thank you.

[Arshia-HZ]

I had the same problem with the detect_video function for YOLOv8! I tried this code for end2end section and problem resolved:

if end2end:
    num, final_boxes, final_scores, final_cls_inds = data

    final_boxes = np.reshape(final_boxes/ratio, (-1, 4))
    # Check if final_scores and final_cls_inds are empty
    if final_scores.size == 0 or final_cls_inds.size == 0:
        print("No detections found or output arrays are empty.")
        dets = final_boxes  # or set to an empty list/array if preferred
    else:
        # Ensure final_scores and final_cls_inds have shape (num_detections, 1)
        final_scores = np.array(final_scores).reshape(-1, 1)  # Shape: (num_detections, 1)
        final_cls_inds = np.array(final_cls_inds).reshape(-1, 1)  # Shape: (num_detections, 1)

        # Ensure that the number of detections is consistent
        if len(final_boxes) != len(final_scores) or len(final_boxes) != len(final_cls_inds):
            print("Mismatch in detection counts!")
            dets = final_boxes  # or set to an empty list/array if preferred
        else:
            # Concatenate final_boxes, final_scores, and final_cls_inds along the last axis
            dets = np.concatenate([final_boxes, final_scores, final_cls_inds], axis=-1)