----------
Boosted Descriptors
----------
This software is an implementation of the boosted descriptors presented in the following papers:
[1] T. Trzcinski, M. Christoudias, P. Fua, V. Lepetit. Boosting Binary Keypoint Descriptors. Computer Vision and Pattern Recognition (CVPR) 2013.
[2] T. Trzcinski, M. Christoudias, V. Lepetit, P. Fua. Learning Image Descriptors with the Boosting-Trick. Neural Information Processing Systems (NIPS) 2012.
The code is released under the GNU General Public License V2 (attached within this package).
COMPILATION:
Type
$ make
to compile the code and generate the executable 'DBRIEF_demo'.
The code depends on OpenCV-2.4 or later, so you should first install
and make sure that pkg-config can locate it. To test it, type
$ pkg-config opencv --libs --cflags
and you should get a list of compiler flags.
Important Notices:
1) This demo relies on Hamming Distance calculation. Some of new
generation processors support SSE4.2 instructions which includes an
instruction called POPCNT which enables fast Hamming Distance
calculation. In g++ SSE4.2 instructions are enabled with -msse4.2
flag. By default, this flag is set in the Makefile. Therefore, if
your instruction set does not support SSE4.2 instructions, although
this code compiles perfectly, it will crash giving an error message.
If you get such an error, please open Makefile with your favorite
editor and remove -msse4.2 flag.
USAGE:
./main (--extract <imgFile> <outputFile> <descriptor>
| --match <imgFile1> <imgFile2> <descriptor>)
Descriptor names: BGM, LBGM, BINBOOST(-64/128/256)
Example usage:
$ ./main --match graf/img1.ppm graf/img2.ppm binboost-256
Image of matched images should be saved to 'matches.png'.
FOR REPORTING BUGS and FEEDBACK:
Please send an email to Tomasz Trzcinski: tomasz.trzcinski@epfl.ch
and/or Mario Christoudias: mario.christoudias@epfl.ch
and/or Vincent Lepetit: vincent.lepetit@epfl.ch.