saeedghsh / Place-Categorization-2D

Semi-Supervised Place Categorization Of 2D Occupancy Grid Maps

Place-Categorization-2D-OGM

Semi-Supervised Place Categorization of Occupancy Grid Maps (OGM) In 2D

This repository conatins a python implementation of place categorization method, explained in this paper (link, pdf):

• Shahbandi, Saeed Gholami, Björn Åstrand, and Roland Philippsen. "Semi-supervised semantic labeling of adaptive cell decomposition maps in well-structured environments." Mobile Robots (ECMR), 2015 European Conference on. IEEE, 2015.

The feature set employed in this work is inspired by (link):

• Oscar Martinez Mozos "Semantic labeling of places with mobile robots" 2010, Springer Berlin Heidelberg.

Image below shows a simple demo of place categorization, with Kmean clustering where number of categories are set to two. For more details and more examples see the abovementioned paper.

Dependencies

To install dependencies:

git clone https://github.com/saeedghsh/Place-Categorization-2D.git
cd Place-Categorization-2D
pip install -r requirements.txt % opencv must be installed separately

Notes

• This repository only contains the core method for place categorization, not the full method from the paper above. For instance:
  • It is not adaptive to environment types, and paramters must be set according to input maps (e.g. resolution). For a better performance, one need to tweak parameters of the clustering algorithm manually (or adaptively).
  • The decomposition of the 2D plane from the abovementioned paper ("Semi-supervised ...") is carried out by the arrangement package.

To be documented soon

• Setting Paramters
• Scripts
• NumPy vectorization VS. multi-processing, and handling memory issue

License

Distributed with a GNU GENERAL PUBLIC LICENSE; see LICENSE.

Copyright (C) Saeed Gholami Shahbandi