This repository corresponds to the work entitled "Unsupervised appearance map abstraction for indoor Visual Place Recognition with mobile robots", published at IEEE Robotics and Automation Letters.
Authors: Alberto Jaenal, Francisco-Angel Moreno and Javier Gonzalez-Jimenez
Video: Click Here
If you use this work in your research, please cite:
@ARTICLE{jaenal2022unsupervised,
author={Jaenal, Alberto and Moreno, Francisco-Angel and Gonzalez-Jimenez, Javier},
journal={IEEE Robotics and Automation Letters},
title={Unsupervised Appearance Map Abstraction for Indoor Visual Place Recognition With Mobile Robots},
year={2022},
volume={},
number={},
pages={1-7},
doi={10.1109/LRA.2022.3186768}}
- Donwload the demo data into the
data
folder:
mkdir data && cd data && wget http://ftp.uma.es/Mapir/MapAbstractionVPR/data/feats-ImRet.npy && wget http://ftp.uma.es/Mapir/MapAbstractionVPR/data/feats-NetVLAD.npy && wget http://ftp.uma.es/Mapir/MapAbstractionVPR/data/poses.txt
- Run the map abstraction script
python run.py -p data/poses.txt -f data/feats-NetVLAD.npy -c 10
This software employs built-in libs (see requeriments.txt
), and has been tested with Python>=3.5 on Ubuntu 16.04, 18.04 and 20.04.
The geometry.py
script is inspired in ProbFiltersVPR.