nobodyinamerica/bold

BOLD - Binary Online Learned Descriptor

BOLD - Binary Online Learned Descriptor

Introduction

Sample implementation of the method described in

@InProceedings{Balntas_2015_CVPR,
author = {Balntas, Vassileios and Tang, Lilian and Mikolajczyk, Krystian},
title = {BOLD - Binary Online Learned Descriptor For Efficient Image Matching},
journal = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2015}
}

Download the paper from http://www.iis.ee.ic.ac.uk/~vb2415/bold/BOLD-Binary-Online-Learned-Descriptor.pdf

Compilation

In order to run the example described in main.cpp you need to download the sample datasets from http://www.iis.ee.ic.ac.uk/~vb2415/bold/dataset-ubc-bold.tar.gz and extract the file inside the main git folder.

To compile the code type

cmake .
make

and run ./bold to run a simple patches experiment on the notredame dataset from from the UBC Patches Dataset

The program will run print a series of evaluation results for the patch pairs from the dataset above eg.

0 296 136
1 125 31
1 237 77
.
.
.
1 147 39
0 292 136
1 125 26

The first column contains the patch pair label (0 negative, 1 positive), the second row contains the unmasked distance between the two patches of the pair, and the third the masked distance using the methodology described in our paper.

Notice that the new masked distance in the case of positive patches is (in terms of ratio) much smaller than the original compared to the negative patches. This confirms the findings that the masked distance is more discriminative.

Contributor needed!

Recent work (PAMI under submission), shows that we can extend the above masking framework to all the other binary descriptors (BRIEF, BRISK, FREAK). If someone is willing to expand the masking framework to the OpenCV class of binary descriptors, send me an email and I can help out.

Note

Currently the BOLD code works with 32x32 patches. In order to generalize it, convert the indexes of the patch tests in the bold.descr file.