When I run demo_camera.py, sometimes the verts or joints3d information of some frames are incorrect,for example,Swap left and right.
However, run demo.py for each image, this will not happen.
Hope to get your explanation, thank you very much!
I think it is just that the input image you're using is a challenging one -- this is still a learning based algorithm and not perfect. There are more recent models, like SPIN, which may be more accurate.
When I run demo_camera.py, sometimes the verts or joints3d information of some frames are incorrect,for example,Swap left and right. However, run demo.py for each image, this will not happen.
Hope to get your explanation, thank you very much!