Skip to content

Latest commit

 

History

History
executable file
·
147 lines (106 loc) · 5.15 KB

README.md

File metadata and controls

executable file
·
147 lines (106 loc) · 5.15 KB
Raw

PSMNet_AIHub

본 저장소는 NIA 인도 보행 공공 데이터의 검수용으로 "Pyramid Stereo Matching Network" 기반으로 구축되었습니다.
인도 보행 데이터는 공공 데이터 구축을 목적으로 하는 AI Hub에서 제공됩니다.
인도 보행 공공 데이터는 장애인 인도보행의 어려움과 이동권 문제 해결을 위하여 만들어졌습니다.

This repository contains the inspection of NIA Sidewalk dataset provided by AI Hub.
Sidewalk dataset is a public dataset to solve that disabled person suffer from the difficulty of mobility in the sidewalk.
This repository contains the code (in PyTorch) for "Pyramid Stereo Matching Network" paper (CVPR 2018) by Jia-Ren Chang and Yong-Sheng Chen.

Introduction

PSMNet, a pyramid stereo matching network, consists two main modules: spatial pyramid pooling and 3D CNN.

Installation

Dependencies

For detailed installation, Please refer INSTALL.md.
자세한 설치과정은 INSTALL_kor.md 을 참고하세요.

Prepare datasets

Dataset Download : https://aihub.or.kr/aidata/136/download
main.py reads images from following data structure. The folder name starting with ZED* and test* is training data and test data respectively.

${datapath}
├── 1
│   ├── ZED1_1
|   │   ├── ZED1_KSC_000000_right.png
|   │   ├── ZED1_KSC_000000_right.png
|   │   ├── ZED1_KSC_000000_right.png
|   │   ├── ...
│   ├── ZED2_1
|   │   ├── ZED2_KSC_000000_right.png
|   │   ├── ZED2_KSC_000000_right.png
|   │   ├── ZED2_KSC_000000_right.png
│   ├── ...
│   ├── test_ZED4_1
|   │   ├── ZED4_KSC_000000_right.png
|   │   ├── ZED4_KSC_000000_right.png
|   │   ├── ZED4_KSC_000000_right.png
|   │   ├── ...
├── ...
├── 7

Execution

sh run.sh

In run.sh, following command are executed. Set --datapath as same as ${datapath} above data structure.

python main.py --maxdisp 192 \
               --model stackhourglass \
               --datapath (your data folder. for example ./NIA/) \
               --epochs 200 \
               --batchsz 12 \
               --nworker 20 \
               --savemodel (path for saving model)/

Pretrained Model

  • NOTE: The pretrained model were saved in .tar; however, you don't need to untar it. Use torch.load() to load it.
  • 학습된 모델이 .tar로 저장되나, 압축을 풀지 말고 torch.load()를 사용하면 됩니다.
KITTI 2015 Scene Flow KITTI 2012 NIA Sidewalk
Google Drive Google Drive Google Drive Google Drive

Results

Hardware

  • 8 NVIDIA Titan Xp GPUs
  • Intel Xeon 4210 CPU @ 2.20GHz

Software

  • Ubuntu 16.04
  • Python 2.7
  • PyTorch 0.4.0
  • CUDA 9.0

Results on NIA Sidewalk dataset

Network Language D1-all (All) Download
PSMNet Pytorch 0.4.0 4.779 % Model

Qualitative results

Left image

Predicted disparity

Contacts

parkkibaek@kaist.ac.kr

License

MIT License

Citation

@inproceedings{chang2018pyramid,
  title={Pyramid Stereo Matching Network},
  author={Chang, Jia-Ren and Chen, Yong-Sheng},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  pages={5410--5418},
  year={2018}
}

SideGuide dataset SideGuide: A Large-scale Sidewalk Dataset for Guiding Impaired People (http://ras.papercept.net/images/temp/IROS/files/1873.pdf)

@inproceedings{park2020sideguide,
  title={SideGuide: A Large-scale Sidewalk Dataset for Guiding Impaired People},
  author={Park, Kibaek and Oh, Youngtaek and Ham, Soomin and Joo, Kyungdon and Kim, Hyokyoung and Kum, Hyoyoung and Kweon, In So},
  booktitle={2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},
  pages={10022--10029},
  year={2020},
  organization={IEEE}
}