search

huiyu-gao / VisFusion

[CVPR 2023] Code for "VisFusion: Visibility-aware Online 3D Scene Reconstruction from Videos"
https://huiyu-gao.github.io/visfusion
Apache License 2.0
181 stars 9 forks source link

Results on 7Scenes Dataset #2

Open UsamaHasan opened 1 year ago

UsamaHasan commented 1 year ago

Hy there, thank you for making your code public and remarkable contribution. I'm trying to recreate your paper result. Thou I was successful in generating them on scannet test dataset but unable to recreate them on 7scene dataset. Is there any specific transformation that you applied for scannet which should be removed or updated in the code? .Also when I change the voxel size the 0.01 in test_config.yaml, the code throughs the following error.

/VisFusion/utils.py", line 130, in update 
self.data[k] += v 
KeyError: 'loss_1'

snapshot00

wei-mao-2019 commented 1 year ago

Hi Usama,

Thank you for your interests.

You need to align the coordinate system first since all the scenes in Scannet used to train the model are well aligned.

The aligned scene should look like this Screenshot from 2023-07-18 12-05-58

Cheers, Wei

UsamaHasan commented 1 year ago

Hi Wei, thanks for replying to this comment. I have tried looking for a solution for that, Can you please share the method or approach you used for coordinate system alignment which you used. It would be of great help.