object_3d_retrieval¶
This contains our implementation of a vocabulary tree (Nister & Stewenius) together with some modifications for our paper. It also includes some functions e.g. for extracting segments & features and querying on our data structure.
To use the system, follow the instructions in scripts/menu.py
. There
are submenus, for example for querying that can be run separately, i.e.
scripts/training_menu.py
and scripts/querying_menu.py
.
Dependencies¶
Right now, we do not use the debian package of strands_3d_mapping
(see below), instead you should compile
https://github.com/RaresAmbrus/strands_3d_mapping manually in a catkin
workspace, be sure to check out the hydro-devel
branch. There is a
commented section in the cmake file where you can set the variable
parser_workspace
, which should point to the catkin workspace where
your strands_3d_mapping
checkout lives. See the lines
https://github.com/nilsbore/dynamic_object_retrieval/blob/dynamic/CMakeLists.txt#L52
and
https://github.com/nilsbore/dynamic_object_retrieval/blob/dynamic/benchmark/CMakeLists.txt#L31.
In the future, you will instead use the packaged version of
strands_3d_mapping
from the STRANDS project
http://strands.acin.tuwien.ac.at/. Follow the instructions on
https://github.com/strands-project-releases/strands-releases/wiki to add
the debian package repository. Then install the metaroom_xml_parser
by typing sudo apt-get install ros-indigo-metaroom-xml-parser
.
Tested with Ubuntu 14.04 with ROS Indigo and corresponding OpenCV + PCL, QT4. The repos https://github.com/nilsbore/k_means_tree, https://github.com/USCiLab/cereal and https://github.com/mp3guy/Stopwatch are included in the repo as subtrees.
Original page: https://github.com/strands-project/strands_3d_mapping/blob/hydro-devel/dynamic_object_retrieval/README.md