search

facebookresearch / banmo

BANMo Building Animatable 3D Neural Models from Many Casual Videos
Other
539 stars 58 forks source link

Question about mesh of Viser in Banmo #43

Closed bravotty closed 2 years ago

bravotty commented 2 years ago

Thank you for your extremely good work. I had a question reproducing viser in that the mesh generated for the cat was smoother but poorly shaped than the mesh you put in the banmo. (Compare below, left is the mesh in banmo, right is the mesh I reran viser to generate) image As well as in ama-female, and the one we generated would tend to have no hands (as shown in the comparison below) image The images above are all n_faces=8000 in the code and are fully trained by the official viser script.

gengshan-y commented 2 years ago

Hi, the comparison with viser in the paper was run with a lower weight on the laplacian smoothness term (triangle_reg_wt=0.005). The latest viser codebase uses triangle_reg_wt=0.025 to produce more visually appealing but smoother result.

You should be able to get the same result as we showed in the paper with triangle_reg_wt=0.005

bravotty commented 2 years ago

Thank you for your reply and I will make the relevant improvements.