Is this detection result a bug or a performance issue?

[VisImage]

I am not sure if the detection result below is a performance issue, or a bug in the configuration. The key points detected from the algorithm is so strange. This result is generated using the following instruction as indicated in the file "https://github.com/open-mmlab/mmpose/blob/master/docs/en/install.md":

Option (b). If you install mmpose with pip, open you python interpreter and copy&paste the following codes.

from mmpose.apis import (init_pose_model, inference_bottom_up_pose_model, vis_pose_result) config_file = 'associative_embedding_hrnet_w32_coco_512x512.py' checkpoint_file = 'hrnet_w32_coco_512x512-bcb8c247_20200816.pth' pose_model = init_pose_model(config_file, checkpoint_file, device='cpu') # or device='cuda:0' image_name = 'demo/1015.jpg' pose_results, _ = inference_bottom_up_pose_model(pose_model, image_name) vis_pose_result(pose_model, image_name, pose_results, out_file='demo/vis_persons.jpg')

original image is as below:

[jin-s13]

It is a performance issue. The bottom-up AE model may not perform very well in crowded scenes. You may also try top-down models. They normally perform better than bottom-up ones.