Binary-feature-trees v0.0.1

Based of the paper "Bags of Binary Words for Fast Place Recognition in Image Sequences, Dorian G ́alvez-L ́opez and Juan D. Tard ́os, Member, IEEE" found here: http://doriangalvez.com/papers/GalvezTRO12.pdf

Instead of using kmean for clustering, I chose to implement kmedoids PAM with k++ seeding.

Installation notes

1. install catkin-tools

sudo apt get install catkin sudo pip3 install git+https://github.com/catkin/catkin_tools.git sudo apt-get install python3-catkin-pkg

2. init catkin workspace

catkin_make

3. catkin config

catkin config --cmake-args -DCMAKE_EXPORT_COMPILE_COMMANDS=ON

4. clean

catkin clean -y

5. build

catkin build

ctrl+p ext install ms-vscode.cpptools ext install betwo.b2-catkin-tools

https://marketplace.visualstudio.com/items?itemName=betwo.b2-catkin-tools

6. catkin build ./devel/lib/trainer/trainer Hello World2

Sanity checker JS script

var results = []; Object.keys(a.clusterMembers).forEach((idx) => { console.log(idx); console.log("x", a.clusterMembers[idx][1].length); Object.keys(a.clusterMembers[idx][1]).forEach((m, memberIdx) => { if (a.clusterMembers[idx][1][memberIdx] == 490) { results.push([idx, memberIdx]); } }); });