search

mvp18 / sp2-360

Code for Sp2360: Sparse-view 360° Scene Reconstruction using Cascaded 2D Diffusion Priors
MIT License
1 stars 0 forks source link

About training view selection #1

Open zhanghaoyu816 opened 3 days ago

zhanghaoyu816 commented 3 days ago

Thanks for your nice project. I want to select sparse training views in some complex scene such as mipnerf360, scannet or tanks&temples so that these views can cover most of the whole scene with less overlap between views. So could you please provide the code for data selection in your project? Thanks a lot.

mvp18 commented 3 days ago

Hi @zhanghaoyu816,

Thanks for your interest in our work. I'll share the view selection code in this thread soon after some cleaning. Meanwhile, you can refer to the ReconFusion splits available here for mipnerf360 splits that maximize scene coverage, but cannot always register all input images to a single point cloud. These splits are more helpful if your 3D representation is NeRF-based and not point-based like 3dGS.

Our splits have more overlap across views but at the cost of lower scene coverage.

Best, mvp18

zhanghaoyu816 commented 3 days ago

Thanks for your reply, I'm looking forward it.