You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When I use the hrnet_w32_h36m_256x256 model and the top_down_img_demo_with_mmdet.py file to inference the h36m data set, the model can give me the correct result, but when I use the model to inference the custom image, the output keypoints score is very low (such as 1.27726495 e-02). May I ask why?
The specific model and code I use are as follows:
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
When I use the hrnet_w32_h36m_256x256 model and the top_down_img_demo_with_mmdet.py file to inference the h36m data set, the model can give me the correct result, but when I use the model to inference the custom image, the output keypoints score is very low (such as 1.27726495 e-02). May I ask why?
The specific model and code I use are as follows:
Beta Was this translation helpful? Give feedback.
All reactions