This repository contains the source code for our paper:
MC-Stereo: Multi-Peak Lookup and Cascade Search Range for Stereo Matching
Miaojie Feng, Junda Cheng, Hao jia, Longliang Liu, Gangwei Xu, Xin Yang
| Method | KITTI-2012 (2-noc) |
KITTI-2012 (2-noc-ref) |
KITTI-2015 (D1-all) |
|---|---|---|---|
| ACVNet | 1.83 | 11.42 | 1.65 |
| RAFT-Stereo | - | - | 1.91 |
| IGEV-Stereo | 1.71 | 7.29 | 1.59 |
| CREStereo | 1.72 | 9.71 | 1.69 |
| MC-Stereo(Ours) | 1.55 | 6.82 | 1.55 |
conda create -n MC_Stereo python=3.7
conda activate MC_Stereo
conda install pytorch==1.10.0 torchvision==0.11.0 torchaudio==0.10.0 cudatoolkit=11.3 -c pytorch -c conda-forge
pip install opencv-python
pip install scikit-image
pip install tensorboard
pip install tqdm
pip install timm==0.5.4
To evaluate/train MC-Stereo, you will need to download the required datasets.
Pretrained models can be downloaded from google drive.
To evaluate on Scene Flow, run
sh evaluate.shTo train on Scene Flow, run
sh train_sceneflow.shTo train on KITTI, run
sh train_kitti.shTo train on ETH3D, run
sh train_eth3d.shFor submission to the KITTI benchmark, run
python save_disp.pyIf you find our work useful in your research, please consider citing our paper:
@article{feng2023mc,
title={MC-Stereo: Multi-peak Lookup and Cascade Search Range for Stereo Matching},
author={Feng, Miaojie and Cheng, Junda and Jia, Hao and Liu, Longliang and Xu, Gangwei and Yang, Xin},
journal={arXiv preprint arXiv:2311.02340},
year={2023}
}Please feel free to contact me (Miaojie) at [email protected].
This project is heavily based on RAFT-Stereo, We thank the original authors for their excellent work.