search

zc-alexfan / hold

[CVPR 2024✨Highlight] Official repository for HOLD, the first method that jointly reconstructs articulated hands and objects from monocular videos without assuming a pre-scanned object template and 3D hand-object training data.
https://zc-alexfan.github.io/hold
MIT License
271 stars 7 forks source link

problem on getting total mesh #13

Closed Lyy-iiis closed 2 months ago

Lyy-iiis commented 2 months ago

How can i get the absolute coordinate of mesh? I try to visualize the object and hand mesh, each single looks well, but when i put them together, the results looks like 5521720774234_ pic which is incorrect because in my original video, hand and object are about same size, how can i find correct coordinate to record the mesh

zc-alexfan commented 2 months ago

what's the name of files that you opened?

Lyy-iiis commented 2 months ago

Under code/logs/$exp_id/mesh_cano, I pick mesh_cano_left_step_misc.obj, mesh_cano_right_step_misc.obj, mesh_cano_object_step_misc.obj out. Then I merge three mesh together, plot them using matplotlib

Lyy-iiis commented 2 months ago

By the way, since the hand is too small, I add a bias to the hand such that you can see it clearly. Without the bias, we can't separate hand and object anymore.

zc-alexfan commented 2 months ago

Those are the canonical shapes, not the posed shapes. You can visualize with visualize_ckpt.py here https://github.com/zc-alexfan/hold/blob/master/docs/usage.md

Lyy-iiis commented 2 months ago

Thank you very much, I find mesh from the visualize_ckpt.py is correct.