kim0930 / manhattan_sdf

Code for "Neural 3D Scene Reconstruction with the Manhattan-world Assumption" CVPR 2022 Oral

News

• 06/03/2022 We provide the instruction to run on custom data here.
• 05/10/2022 To make the comparison on ScanNet easier, we provide all quantitative and qualitative results of baselines here, including COLMAP, COLMAP*, ACMP, NeRF, UNISURF, NeuS, and VolSDF.
• 05/10/2022 To make the following works easier to compare with our model, we provide our quantitative and qualitative results, as well as the trained models on ScanNet here.
• 05/10/2022 We upload our processed ScanNet scene data to Onedrive.

Neural 3D Scene Reconstruction with the Manhattan-world Assumption

Project Page | Video | Paper

Neural 3D Scene Reconstruction with the Manhattan-world Assumption
Haoyu Guo^*, Sida Peng^*, Haotong Lin, Qianqian Wang, Guofeng Zhang, Hujun Bao, Xiaowei Zhou
CVPR 2022 (Oral Presentation)

Setup

Installation

conda env create -f environment.yml
conda activate manhattan

Data preparation

Download ScanNet scene data evaluated in the paper from Onedrive / Google Drive / BaiduNetDisk (password:ap9k) and extract them into data/. Make sure that the path is consistent with config file.

We provide the instruction to run on custom data here.

Usage

Training

python train_net.py --cfg_file configs/scannet/0050.yaml gpus 0, exp_name scannet_0050

Mesh extraction

python run.py --type mesh_extract --output_mesh result.obj --cfg_file configs/scannet/0050.yaml gpus 0, exp_name scannet_0050

Evaluation

python run.py --type evaluate --cfg_file configs/scannet/0050.yaml gpus 0, exp_name scannet_0050

Citation

If you find this code useful for your research, please use the following BibTeX entry.

@inproceedings{guo2022manhattan,
  title={Neural 3D Scene Reconstruction with the Manhattan-world Assumption},
  author={Guo, Haoyu and Peng, Sida and Lin, Haotong and Wang, Qianqian and Zhang, Guofeng and Bao, Hujun and Zhou, Xiaowei},
  booktitle={CVPR},
  year={2022}
}

Acknowledgement

• Thanks to Lior Yariv for her excellent work VolSDF.
• Thanks to Jianfei Guo for his implementation of VolSDF neurecon.
• Thanks to Johannes Schönberger for his excellent work COLMAP.
• Thanks to Shaohui Liu for his customized implementation of COLMAP as a submodule of NerfingMVS.