yanghb22-fdu/Hi3D-Official

3D Mesh Extraction

Opened this issue · 1 comments

wondering516 commented

Thanks for your great works, I am still confused about the part 4.4 in the paper.

  • As described in part 4.4, between frameA and frameB you using 3DGS to get more novel views, how many frames in total are enough for SDF-based reconstruction method to extract 3D mesh

  • Is there any possibility for authors to release the code of part 4.4

Many Thanks for authors.

2hiTee commented

I have the same question, the author says first leverage 3DGS to reconstruct the object and render more views. But the paper says it just produce 16 views from the diffusion model, which I think it is not enough to run COLMAP to get the correct camera pose for 3DGS reconstruction.