IndexError: list index out of range

[spurani]

Hi, Here is the command which I am executing for real time object detection using jetson tx2 onboard camera. After I execute this command I can see the my tf-trt graph.pb it loads successfully then I can see my webcam output which is inverted and after few seconds it freezes and throws the errror below. Can you please guide me through this so that i can continue learning these concepts and understand it? I have exactly 3 classes in my label file

python3 camera_tf_trt.py --latency 200 --model new_model_ssd_colab_v2_mobilenet_2020_29_12 --labelmap data/label_final.pbtxt --num-classes 3 --build

INFO:main:warming up the TRT graph with a dummy image INFO:main:starting to loop and detect Gtk-Message: 02:24:07.146: Failed to load module "canberra-gtk-module" Traceback (most recent call last): File "camera_tf_trt.py", line 198, in main() File "camera_tf_trt.py", line 188, in main loop_and_detect(cam, tf_sess, args.conf_th, vis, od_type=od_type) File "camera_tf_trt.py", line 120, in loop_and_detect img = vis.draw_bboxes(img, box, conf, cls) File "/home/s/projects/tf_trt_models/utils/visualization.py", line 96, in draw_bboxes color = self.colors[cl] IndexError: list index out of range Killed Thank you really great tutorial I am learning a lot. Please let me know if you require more details

[jkjung-avt]

This is likely because: (a) you have specified --num-classes 3, (b) but the model actually outputs a class id which is greater than or equal to 3.

[spurani]

{0: 'CLS0', 1: 'bottle', 2: 'cardboard', 3: 'clothing'} this is the list of the classes in cls_dict so I changed --num-classes 4 still it didn't work. I am not sure about 0: 'CLS0' is that for background? I can see this class gets added here along with the classes from label file /home/s/projects/tf_trt_models/utils/od_utils.py

def read_label_map(path_to_labels):
    """Read from the label map file and return a class dictionary which
    maps class id (int) to the corresponding display name (string).

    Reference:
    https://github.com/tensorflow/models/blob/master/research/object_detection/object_detection_tutorial.ipynb
    """
    from object_detection.utils import label_map_util

    category_index = label_map_util.create_category_index_from_labelmap(
        path_to_labels)
    cls_dict = {int(x['id']): x['name'] for x in category_index.values()}
    num_classes = max(c for c in cls_dict.keys()) + 1
    # add missing classes as, say,'CLS12' if any
    return {i: cls_dict.get(i, 'CLS{}'.format(i)) for i in range(num_classes)}

Can you please guide me through this? thank you

[jkjung-avt]

Sorry, the "--num_classes" command-line option is deprecated and not used. The code actually determines number of object classes by reading the labelmap file.

{0: 'CLS0', 1: 'bottle', 2: 'cardboard', 3: 'clothing'} this is the list of the classes in cls_dict so I changed --num-classes 4 still it didn't work. I am not sure about 0: 'CLS0' is that for background?

Yes, class id 0 of SSD models usually refers to the "background" class.

To debug the issue, please add a print() statement (or use pdb) to see what cls value your model outputs. For example, add a print() after this line.

    if cls > 3:  print('ERROR: cls id is ', cls)

[spurani]

Hi I am getting this error

Gtk-Message: 15:23:01.286: Failed to load module "canberra-gtk-module"
Traceback (most recent call last):
  File "camera_tf_trt.py", line 202, in <module>
    main()
  File "camera_tf_trt.py", line 192, in main
    loop_and_detect(cam, tf_sess, args.conf_th, vis, od_type=od_type)
  File "camera_tf_trt.py", line 126, in loop_and_detect
    cv2.imshow(WINDOW_NAME, img)
cv2.error: OpenCV(4.1.1) /home/nvidia/host/build_opencv/nv_opencv/modules/highgui/src/window.cpp:352: error: (-215:Assertion failed) size.width>0 && size.height>0 in function 'imshow'

Can you please guide me how to fix it?

[jkjung-avt]

cv2.error: OpenCV(4.1.1) /home/nvidia/host/build_opencv/nv_opencv/modules/highgui/src/window.cpp:352: error: (-215:Assertion failed) size.width>0 && size.height>0 in function 'imshow'

This error message indicates the image you called cv2.imshow() with is not valid. You could use, say, pdb to check what might have gone wrong.