Inference video in the wild but got wrong result

Question

Inference video in the wild but got wrong result

Opened this issue 3 years ago · 3 comments

z0978916348 commented 3 years ago

Hi,

I used HRNet to generate 2d keypoint npz follow VideoPosed3D format.

I slightly modified your code to inference video in the wild follow VideoPose3D but got the wrong result as below:

The result is all entangled.

Can you point out my mistake?

Thanks

Answer 1 · 2021-12-06T02:07:48.000Z

Would you please share the modified code?

Answer 2 · 2022-03-30T09:27:24.000Z

I also get similar results like yours.

Answer 3 · 2022-04-10T08:53:11.000Z

Sorry that iI can't solve your problem, but can you tell me how exactly did you get this video (Should I install a 2D detector and and use its output as the input of poseformer) ?