dazinovic / neural-rgbd-surface-reconstruction

Official implementation of the CVPR 2022 Paper "Neural RGB-D Surface Reconstruction"

Home Page:https://dazinovic.github.io/neural-rgbd-surface-reconstruction/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

release scannet experiment data

Guptajakala opened this issue · comments

commented

Hi, thanks for sharing this fantastic work!
Would you have any plan to release the processed scanNet data for the experiments? Especially for the computed camera params, train/val pose split and filtered depth images, which could be good starting point for others to compare fairly.

commented

Hi Guptajakala,

I do not plan to release any ScanNet data (please refer to the official ScanNet website to obtain the data). There is no train/val split here. Every 'train_skip'th pose (e.g. every pose or every 4th pose) is used to optimize the SDF. The trainval_poses.txt naming scheme is an artifact from an earlier stage in the project and I forgot to rename it for the code release.

To get the filtered depth images, you can run https://github.com/angeladai/spsg/blob/master/datagen/src/CameraUtil.h#L25 with sigmaD = 2.0 and sigmaR = 0.1.

If you need help with comparisons for your paper (e.g. comparison of the camera trajectories), feel free to send me an email.

Can you guys please help me to run this code on ScanNet? Struggling for last two weeks.

@soumyadipm Hi, did you manage to run on ScanNet? Help is appreciated.