icra2018/maoris repository overview

⁠MAORIS : MAp Of RIpples' Segmentation

Cutting most types of maps in smaller more meaningful bits.

⁠Description

This program segment maps in semantically meaningful parts, such as rooms and corridors. It has been tested on robot build maps and sketch maps, but the principle can easily be extended to other type of maps such as city maps.

It also provides a way to evaluate and test your own segmentation using Matthews Correlation Coefficient⁠.

A little bit more about this here⁠.

⁠Paper

The paper describing the method was accepted for publication at ICRA2017 (:D!) and should be published soon after the conference. The preprint version is available on Arxiv⁠.

⁠How to compile

⁠Dependencies

• boost
• openCV 2.9 or more (openCV 3 included)
• bettergraph⁠
• VoDiGrEx⁠

Compile by doing

mkdir release
cmake ..
make

If one need to use maoris for other project you can do

sudo make install

to install and

sudo make uninstall

To uninstall.

⁠How to use maoris

After compilation, you will have those executables:

⁠For segmentation:

• segmentation_live : use the webcam of your laptop to segment maps. Just take a picture of the map and its ground truth and there you go.

• evaluation_2files_fodler : compare segmented images files from two folders

• evaluation_gt : compare a map with a segmented ground truth image

⁠For evaluation

• evaluation_2gt : compare two segmented images

• evaluation_gt_all_files : segment all maps in a folder and then compare them to all segmented ground truths in another folder This file returns a gnuplot dat file named maoris_all_measures.dat with all the data from the segmentation.

• evaluation_gt_param : evaluate maoris' parameters This file returns four gnuplot dat files named maoris_param_threshold.dat, maoris_param_margin.dat, maoris_param_ripples.dat, and maoris_param_doors.dat, with all the data from the segmentation for each parameters.

The arguments of all programs are given in the command line as in evaluation_gt_all_files folder1 folder2 should_draw.

⁠Evaluating your own segmentations

You can evaluate your own segmentation results to ground truths by doing:

evaluation_2files_fodler folder_with_your_segmentation folder_with_your_ground_truth

We also provide segmentation by maoris, DuDe and a Voronoi based segmentation. DuDe segmentation results are in Test/RobotMap/DuDe_results for the robot maps and Test/Sketches/DuDe for the sketch maps. The Voronoi based segmentation results are in Test/RobotMap/Voronoi for the robot maps and Test/Sketches/Voronoi for the sketch maps. maoris segmentation results are in Test/RobotMap/maoris for the robot maps and Test/Sketches/maoris for the sketch maps.

⁠How to Run with Docker

⁠Linux

Tested on Ubuntu 16.04.6 with Docker 18.06.1-ce.

1. Open a terminal and run the command:

docker run --rm -p 8888:8888 -e DISPLAY -v /tmp/.X11-unix:/tmp/.X11-unix  icra2018/maoris:latest

1. Run a web browser and open the link: http://localhost:8888/lab/tree/README.ipynb⁠

⁠macOS

⁠Prerequisites

• XQuartz⁠:
  • Activate the option Allow connections from network clients in XQuartz settings
  • Quit & restart XQuartz (to activate the setting)
  • Open a terminal and run the command:
    xhost + localhost

Tested on macOS Mojave 10.14.3 with XQuartz 2.7.11 and Docker Desktop 2.0.0.3 (engine: 18.09.2).

1. Open a terminal and run the command:

docker run --rm -p 8888:8888 -e DISPLAY=host.docker.internal:0 -v /tmp/.X11-unix:/tmp/.X11-unix  icra2018/maoris:latest

1. Run a web browser and open the link: http://localhost:8888/lab/tree/README.ipynb⁠

⁠Windows

⁠Prerequisites

• VcXsrv⁠ or another X Windows Server
  • Run Xlaunch from the Start menu
  • Select Multiple windows, Start no client, and set Disable access control

Tested on Windows 10 Education wih VcXsrv 1.20.1.4 and Docker Desktop 2.0.0.3 (engine: 18.09.2).

1. Open a Windows PowerShell and run the command:

docker run --rm -p 8888:8888 -e DISPLAY=host.docker.internal:0  icra2018/maoris:latest

1. Run a web browser and open the link: http://localhost:8888/lab/tree/README.ipynb⁠