https://github.com/ducha-aiki/wxbs-descriptors-benchmark

W1BS local patch descriptors benchmark
https://github.com/ducha-aiki/wxbs-descriptors-benchmark

benchmark descriptors descriptors-benchmark image-matching local-features sift wxbs

Last synced: about 1 year ago
JSON representation

W1BS local patch descriptors benchmark

• Host: GitHub
• URL: https://github.com/ducha-aiki/wxbs-descriptors-benchmark
• Owner: ducha-aiki
• Created: 2017-04-07T13:31:15.000Z (about 9 years ago)
• Default Branch: master
• Last Pushed: 2018-10-27T15:23:58.000Z (over 7 years ago)
• Last Synced: 2025-03-11T10:38:06.771Z (about 1 year ago)
• Topics: benchmark, descriptors, descriptors-benchmark, image-matching, local-features, sift, wxbs
• Language: Python
• Homepage: http://cmp.felk.cvut.cz/wbs/
• Size: 24.4 KB
• Stars: 9
• Watchers: 4
• Forks: 4
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
This is W1BS descriptors benchmark from paper [WxBS: Wide Baseline Stereo Generalizations](https://arxiv.org/abs/1504.06603) .

Dataset format:

    data/W1BS/ - directories with subsets. G - geometry, A - appearance, S - sensor, map2photo - map vs. photo

    Each directory contains: 1: regerence image dir, 2 - "noised" image dir, h - homography 1to2 dir

    each image dir contains several images, e.g. dir (data/W1BS/G/1) = 

    [arch.keys  obama.keys  vprice0.keys  vprice1.keys  vprice2.keys  yosemite.keys

    arch.png   obama.png   vprice0.png   vprice1.png   vprice2.png   yosemite.png]

    *.png = image, *.keys = text file with affine keypoints in format: 

        npoints

        x y 5.192*s a11 a12 a21 a22

    *.bmp - hpatches-style column image with pre-extracted patches

    

How to get example results (for now, SIFT, BRIEF and ResizeTo11x11 descriptors are available ): 

    cd data

    ./download_W1BS_dataset.sh

    cd ../code

    ./do_everything.sh

To add your descriptor to benchmark, please add corresponding script to code/descriptors directory.

The provided file should take two arguments: path to input image input_img.bmp and path to output text file with descriptors. 

Output file: one space separated line for one descriptor.

Please, see example in [code/descriptors/Pixels11.py](https://github.com/ducha-aiki/wxbs-descriptors-benchmark/blob/master/code/descriptors/Pixels11.py) or [code/descriptors/SIFT.py](https://github.com/ducha-aiki/wxbs-descriptors-benchmark/blob/master/code/descriptors/SIFT.py)

## Citation

Please cite us if you use this code:

    @InProceedings{Mishkin2015WXBS,

     author = {{Mishkin}, D. and {Matas}, J. and {Perdoch}, M. and {Lenc}, K. },

     Booktitle = {Proceedings of the British Machine Vision Conference},

     Publisher                = {BMVA},

     title = "{WxBS: Wide Baseline Stereo Generalizations}",

     year = 2015,

     month = sep,}