NekSfyris / fbow_ros

Fast-bag-of-words ROS wrapper for feature-based place detection and relocalization.

fbow_ros

Fast-bag-of-words ROS wrapper.

Prerequisites

To use this package you need to have:

• Ubuntu 16.04 LTS
• OpenCV 3.3
• ROS Kinetic

Setup

To use this package you first have to install the main fbow package. To do so, do:

cd ~/catkin_ws/src
git clone https://github.com/rmsalinas/fbow
cd fbow
mkdir build && cd build
cmake ..
make
sudo make install

Then clone the wrapper and build it in your catkin workspace:

cd ~/catkin_ws/src
git clone https://github.com/NekSfyris/fbow_ros
cd ..
catkin_make

Vocabulary

To create your own vocabulary, you have to create a database of images from which the features in the vocabulary will be extracted.

In my case, i created a rosbag from a D435 realsense feed (coloured) of an environment, and then converted it to images. The database of images was kept at the following folder "~/catkin_ws/src/fbow_ros/voc_images/1".

Then, to create the actual vocabulary we first go to the build folder of the fbow package:

cd ~/catkin_ws/src/fbow/build/utils

At last, we run the scripts that do the actual work (check the scripts for more info on the parameters):

./fbow_create_voc_step0 orb ~/Desktop/my_voc ~/catkin_ws/src/fbow_ros/voc_images/1 
./fbow_create_voc_step1 ~/Desktop/my_voc ~/Desktop/voc.fbow -k 10 -l 6 -t 12