website | arxiv | official_repo |
An unofficial and improved implementation of "NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction".
@inproceedings{wang2021neus,
title={NeuS: Learning Neural Implicit Surfaces by Volume Rendering for Multi-view Reconstruction},
author={Wang, Peng and Liu, Lingjie and Liu, Yuan and Theobalt, Christian and Komura, Taku and Wang, Wenping},
booktitle={Proc. Advances in Neural Information Processing Systems (NeurIPS)},
volume={34},
pages={27171--27183},
year={2021}
}instant_neus_gundam_nvs_camera_1l.mp4
instant_neus_stump_nvs_ds.1.0_camera_1l.mp4
instant_neus_bicycle_nvs_ds.1.0_camera_1l.mp4
- Stable training within 10 minutes without necessarily needing mask
- Worried about your camera pose accuracy ? We can refine them !
- Worried about your footage quality & consistency ? We have in the wild image embeddings !
- Worried about geometric distortions like depressions or bulges ? We opt to use monocular normal priors !
- Object-centric, indoor or outdoors ? We can cover them all !
- <10 mins training time on single RTX3090
- 6 GiB GPU Mem
| Dataset | Config file |
|---|---|
| COLMAP dataset + Apperance Embeddings | lotd_neus.colmap.230826.yaml |
| COLMAP dataset + Pose refinement + Apperance Embeddings | lotd_neus.colmap_refine.230826.yaml |
| BlendedMVS dataset preparation | lotd_neus.bmvs.230814.yaml |
| NeuS/DTU dataset preparation |
For detailed instructions, please refer to the general guide section in code_single.
Can be viewed as an unofficial and improved implementation of "MonoSDF: Exploring Monocular Geometric Cues for Neural Implicit Surface Reconstruction".
website | arxiv | offcial_repo |
@inproceedings{Yu2022MonoSDF,
author = {Yu, Zehao and Peng, Songyou and Niemeyer, Michael and Sattler, Torsten and Geiger, Andreas},
title = {MonoSDF: Exploring Monocular Geometric Cues for Neural Implicit Surface Reconstruction},
booktitle={Proc. Advances in Neural Information Processing Systems (NeurIPS)},
year = {2022},
}- <10 mins training time on single RTX3090
- 6 GiB GPU Mem
Follow this link to download the MonoSDF's preprocessed data of Replica / scannet indoor datasets.
| Settings / Dataset | Config file |
|---|---|
| Replica dataset (processed by MonoSDF) | lotd_neus.replica.230814.yaml |
| Scan net dataset (processed by MonoSDF) | WIP |
For detailed instructions, please refer to the general guide section in code_single.