search

zju3dv / manhattan_sdf

Code for "Neural 3D Scene Reconstruction with the Manhattan-world Assumption" CVPR 2022 Oral
https://zju3dv.github.io/manhattan_sdf/
Other
491 stars 35 forks source link

Scannet scenes used for training #24

Closed yinlily12 closed 1 year ago

yinlily12 commented 1 year ago

Hi, thanks for sharing the amazing work!

I am curious about the number of scenes you used for training for the reconstruction task on the Scannet and 7-Scenes datasets . Did you train the model on all scenes or just a few scenes for the reconstruction?

Looking forward to your reply.

ghy0324 commented 1 year ago

Hi, thanks for your interest! Our method perform reconstruction by per-scene optimization, which is similar to NeRF, VolSDF and so on. So we do not train on some scenes for reconstruction task.