Questions about different versions of mmpose

[Deep-learning999]

When I run a cell

Visualize Skeleton vis_frames = vis_skeleton(vid_path, anno, gym_categories[anno['label']]) vid = mpy.ImageSequenceClip(vis_frames, fps=24) vid.ipython_display()

in mmaction2/demo/visualize_heatmap_volume.ipynb, it raise an 'devided by zero' error, I can run it correctly before.
I check the traceback, and I remember I changed my mmpose version from 0.13.0 to 0.16.0 to use the latest posec3d demo.(Posec3d demo can't use with the version of mmpose under 0.16.0)

How to solve this problem

[jin-s13]

Please also raise an issue in mmaction2.

[Deep-learning999]

There are already similar problems above and have not been solved yet
open-mmlab/mmaction2#1029

[jin-s13]

@kennymckormick

[ly015]

Could you please upload full error information to help us locate the problem? @Deep-learning999

[Deep-learning999]

​

TypeError Traceback (most recent call last)
/usr/local/lib/python3.8/dist-packages/numpy/lib/shape_base.py in split(ary, indices_or_sections, axis)
866 try:
--> 867 len(indices_or_sections)
868 except TypeError:

TypeError: object of type 'int' has no len()

During handling of the above exception, another exception occurred:

ZeroDivisionError Traceback (most recent call last)
in
1 # Visualize Skeleton
----> 2 vis_frames = vis_skeleton(vid_path, anno, gym_categories[anno['label']])
3 #vid = mpy.ImageSequenceClip(vis_frames, fps=24)
4 #vid.ipython_display()

in vis_skeleton(vid_path, anno, category_name, ratio)
56 bbox = np.zeros([0, 4], dtype=np.float32)
57 result = [dict(bbox=bbox, keypoints=k) for k in kp]
---> 58 vis_frame = vis_pose_result(model, f, result)
59
60 if category_name is not None:

mmpose/mmpose/apis/inference.py in vis_pose_result(model, img, result, radius, thickness, kpt_score_thr, bbox_color, dataset, show, out_file)
748 raise NotImplementedError()
749
--> 750 img = model.show_result(
751 img,
752 result,

/mmpose/mmpose/models/detectors/top_down.py in show_result(self, img, result, skeleton, kpt_score_thr, bbox_color, pose_kpt_color, pose_limb_color, text_color, radius, thickness, font_scale, bbox_thickness, win_name, show, show_keypoint_weight, wait_time, out_file)
279 labels = [res['label'] for res in result]
280 # draw bounding boxes
--> 281 imshow_bboxes(
282 img,
283 bboxes,

/mmpose/mmpose/core/visualization/image.py in imshow_bboxes(img, bboxes, labels, colors, text_color, thickness, font_scale, show, win_name, wait_time, out_file)
40
41 # adapt to mmcv.imshow_bboxes input format
---> 42 bboxes = np.split(bboxes, bboxes.shape[0], axis=0)
43 if not isinstance(colors, list):
44 colors = [colors for _ in range(len(bboxes))]

<array_function internals> in split(*args, **kwargs)

/usr/local/lib/python3.8/dist-packages/numpy/lib/shape_base.py in split(ary, indices_or_sections, axis)
869 sections = indices_or_sections
870 N = ary.shape[axis]
--> 871 if N % sections:
872 raise ValueError(
873 'array split does not result in an equal division') from None

ZeroDivisionError: integer division or modulo by zero

[Deep-learning999]

I want to install the latest version of mmpose and mmaction2 will have this problem

[ly015]

mmpose v0.13.0 uses imshow_bbox from mmcv. mmpose v0.16.0 switched to imshow_bbox implemented in mmpose for better visualization effect, which failed to handle the case of empty input bbox. This will be fixed soon.

[kennymckormick]

We will try to fix that problem in mmaction2.

[Deep-learning999]

Okay thank you