This is the official source code of the paper 'Rethinking Disparity: A Depth Range Free Multi-View Stereo Based on Disparity.'
Please follow the instruction from IterMVS.
please check the 'train_blend_aug.sh' and 'train_dtu_aug.sh' and update the 'data_root' to your folder.
Please check the 'eval_dtu.sh' and 'eval_tanks.sh' to generate point clouds and update the 'outdir' to your folder.
As for evaluation for DTU, please run the Matlab code under 'evaluations/dtu'. The results look like this:
Acc. (mm) | Comp. (mm) | Overall (mm) |
---|---|---|
0.354 | 0.324 | 0.339 |
Please download pretrained models in this link: https://drive.google.com/drive/folders/1V3JfsZiJunqqKKcHrmk3q5-ybfapKGrq?usp=drive_link
more guide
If you find this project useful for your research, please cite the following:
@article{yan2022rethinking,
title={Rethinking Disparity: A Depth Range Free Multi-View Stereo Based on Disparity},
author={Yan, Qingsong and Wang, Qiang and Zhao, Kaiyong and Li, Bo and Chu, Xiaowen and Deng, Fei},
journal={arXiv preprint arXiv:2211.16905},
year={2022}
}