OmniMVS

This repository contains python codes for paper, "End-to-End Learning for Omnidirectional Stereo Matching with Uncertainty Prior" (TPAMI).

Contact: Changhee Won (changhee.1.won@gmail.com)

Prerequisites

List of code/library dependencies

• Pytorch (tested on 1.5.1)
• pip install numpy scipy matplotlib pyyaml EasyDict scikit-image

How to run

Test (run_test_omnimvs.py)

• Pretrained weights:
  • [omnimvs_plus_ft.pt]
  • [tiny_plus_ft.pt]
• Set arguments in the script
• Run python run_test_omnimvs.py [path_to_pt_file] [dbname]

Example results

Dataset

You can download the synthetic datasets in the project page.

The directory structure should be like this:

[db_root]/[dbname]/[cam%d]/[%05d.png]
                  /omnidepth_gt_640/%05d.tiff  # not necessary
                  /config.yaml
                   ...
(e.g.)
data/sunny/
          /cam1/
          /cam2/
          /cam3/
          /cam4/
          /omnidepth_gt_640/
          /config.yaml

Citation & Acknowledgement

We founded a start-up company MultiplEYE co. ltd. based on this research.

@article{won2020end,
    title={End-to-End Learning for Omnidirectional Stereo Matching with Uncertainty Prior},
    author={Won, Changhee and Ryu, Jongbin and Lim, Jongwoo},
    journal={IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)},
    year={2020},
}