cvlab-stonybrook / s-volsdf

Official implementation of "S-VolSDF: Sparse Multi-View Stereo Regularization of Neural Implicit Surfaces" (ICCV 2023)
https://hao-yu-wu.github.io/s-volsdf/
MIT License
29 stars 1 forks source link

How to reconstruct other scenes on DTU or BMVS #6

Open yec22 opened 7 months ago

yec22 commented 7 months ago

Hi, thanks for the great work!

I'm curious about how to reconstruct other scenes on DTU or BMVS. Can you kindly provide the code that you pre-process the data (especially how to get the 'eval_mask' and 'mvs_data')

Thanks, Looking forward to your reply.

hao-yu-wu commented 7 months ago

Hi, some annotations in eval_mask come from IDR, and some are created by myself with remove.bg. For cam.txt files in mvs_data, I simply copied them from DTU MVS repository and BlendedMVS repository. Also, cameras.npz files are copied from VolSDF. VolSDF did some post-processing to cam.txt, so cam.txt and cameras.npz are in fact consistent with each other.

yec22 commented 7 months ago

Hi, thanks for the reply. How about the 'pair.txt' in 'mvs_data'. I didn't find those files in 'DTU MVS repository'. And, if I want to run S-VolSDF on my custom data, how can I get the 'pair.txt' files.

hao-yu-wu commented 7 months ago

Hi, I forgot to mention that I used pre-processed mvsdata (cam.txt, pair.txt) from TransMVSNet: "For DTU testing set, you can download the preprocessed [DTU testing data](https://drive.google.com/open?id=135oKPefcPTsdtLRzoDAQtPpHuoIrpRI) (from Original MVSNet) and unzip it as the test data folder, which should contain one cams folder, one images folder and one pair.txt file."

If you have custom data, check this. It's COLMAP_reconstruction to mvs_data (cam.txt, pair.txt). Then, apply this to get cameras.npz

yec22 commented 7 months ago

Thanks. I'll try.

14Marchz commented 1 week ago

Thanks. I'll try.

Hello! Could you please share the results of your tests in other DTU scenarios? Thank you.