FBOW (Fast Bag of Words) is an extremmely optimized version of the DBow2/DBow3 libraries. The library is highly optimized to speed up the Bag of Words creation using AVX,SSE and MMX instructions. In loading a vocabulary, fbow is ~80x faster than DBOW2 (see tests directory and try). In transforming an image into a bag of words using on machines with AVX instructions, it is ~6.4x faster in the ORB vocabulary of ORB2Slam.
* Only depends on OpenCV
* Any type of descriptors allowed out of the box (binary and real)
* Dictionary creation from a set of images. Bugs found in DBOW2/3 corrected.
* Extremmely fast bow creation using specialized versions using AVX,SSE and MMX instructions both for binary and floating point descriptors.
* Very fast load of vocabularies
* Not yet implemented indexing of images.
If use this project please cite
@online{FBow, author = {Rafael Muñoz-Salinas},
title = {{FBow} Fast Bag of Words},
year = 2017,
url = {https://github.com/rmsalinas/fbow},
urldate = {2017-02-17}
}
In directory vocabularies you have the ORBSLAM2 vocabulary (https://github.com/raulmur/ORB_SLAM2/tree/master/Vocabulary) in fbow format.
Go to test and run the program test_dbow2VSfbow. Fbow is ~80x faster than DBOW2 in loading the vocabulary, and ~6.4x in transforming an image into a bag of words.