re_indirect_search

Indirect Object Search Algorithm for RoboEarth final demonstrator.

Installation

Before rosmake you may nedd to install the python module scikit-learn

sudo apt-get install build-essential python-dev python-numpy python-setuptools python-scipy libatlas-dev libatlas3-base
sudo pip install -U scikit-learn 
sudo pip install -U jsonpickle
sudo pip install simplejson

###Usage

Batch Learning

• Learn all GMM models (and go for a coffee)

  rosrun re_indirect_search learner.py learn
  

  Note that the learning has to be done when you install the package for the first time

• Get a dump of all the GMM models

  rosrun re_indirect_search learner.py dump <filename>
  

Inference

• Run the indirect_search node

  rosrun re_indirect_search indirect_search.py
  

• Test by running

  cd tests/
  ./sample-inference.py
  

  This script uses the InferenceQuery.srv defined in this package. See script for more details.

One Shot Learning (For RoboEarth Demos Only)

• Run the indirect_search node

  rosrun re_indirect_search indirect_search.py
  

• Test by running

  cd tests/
  ./single-sample-learning.py
  

  This script uses the LearnQuery.srv defined in this package. See script for more details.

Uploading to RoboEarth DB

• The following command publishes the SemMap.msg topics to be uploaded to the RoboEarth (~1449 SemMaps)

  rosrun re_indirect_search uploader.py <optional time[s] to sleep in between>
  

  Default value for the time option 1.0 [s]

Notes for RoboEarth Final Demonstrator

• For naming convensions see data/objectDefinitions.txt. This file has a one-to-one mapping between KnowRob definitions and the names used in the NYU dataset.