search

zju3dv / EasyMocap

Make human motion capture easier.
Other
3.64k stars 449 forks source link

General Question - 3D Computer Vision Tasks in zju3dv lab #335

Open Dipankar1997161 opened 1 year ago

Dipankar1997161 commented 1 year ago

Hello @chingswy

This is a different question which is not part of Easymocap, however, I believe you could respond.

I have been assigned another project, where I need to do the same 3D Reconstruction but for Animals, particularly Deer. So, I was wondering, if "zju3dv Lab" supports any such operations regarding animal reconstruction. It will be helpful.

In case you are aware of any other source or lab, kindly inform me. I look forward to hearing.

Thank you once again

chingswy commented 1 year ago

I believe this issue is relatively straightforward. Once you implement keypoint detection for animals, parameterize the model fitting, and ensure compatibility with the neural body, you can refer to the relevant paper "Artemis: Articulated Neural Pets with Appearance and Motion Synthesis" for further guidance.

Dipankar1997161 commented 1 year ago

I believe this issue is relatively straightforward. Once you implement keypoint detection for animals, parameterize the model fitting, and ensure compatibility with the neural body, you can refer to the relevant paper "Artemis: Articulated Neural Pets with Appearance and Motion Synthesis" for further guidance.

Artemis follows multi-view inputs. But the Data I have is monocular. Any approach now?

chingswy commented 1 year ago

There exists some related works: BITE: Beyond Priors for Improved ThreeD Dog Pose Estimation

Dipankar1997161 commented 1 year ago

Yes, I saw BITE. But unfortunately the code is yet to be made available( publicly)

Also I have the keypoints, so is there a way to generate the SMPL for animals ( model fitting ). Do let me know