eglrp / v4r

This is the github mirror of https://rgit.acin.tuwien.ac.at/v4r/v4r used to provide a public issue board on github.

The V4R library is an open source project from the Vision4Robotics (V4R) group at the ACIN institute, TU Vienna. It contains various state-of-the-art Computer Vision methods and applications especially designed for 3D point cloud processing with a strong focus on robotics applications.

Dependencies:

V4R is stronlgy based on RGB-D data and uses OpenCV and PCL to process data. Here is a list of required components:

Name	Version	Licence
OpenCV	2.4+ or 3.x	3-clause BSD
Eigen	3+	MPL2
Boost	1.48+	Boost Software License
PCL	1.7.2 or 1.8.x	BSD
glog		New BSD
CMake	3.5.1+	3-clause BSD

This is a list of optional components:

Name	Version	Licence	required[r] or optionally-used[o] by following V4R modules	Comments
Ceres*	1.13	New BSD	reconstruction[r], camera_tracking_and_mapping[r], registration[r], recognition[o]
LibSVM	3.20	modified BSD	ml[o]
SiftGPU*	v400	Copyright (c) 2007 University of North Carolina at Chapel Hill (see license)	features[o]	required if SIFT is enabled; can be replaced by non-free OpenCV SIFT implementation. This option is enabled if BUILD_SIFTGPU is disabled in cmake.
EDT*		BSD	features, recognition	computes the Euclidean Distance Transform for point clouds
openNURBS*		No restrictions. Neither copyright nor copyleft restrictions apply.	attention_segmentation[r]	to model smooth surfaces
Caffe		2-Clause BSD	features[o]	Deep Learning Framework used for feature extraction.
OpenGL		Software developers do not need to license OpenGL to use it in their applications. They can simply link to a library provided by a hardware vendor.	rendering[r]
Assimp		BSD	rendering[r]	used to import 3D shape models for various common file types
GLEW		Modified BSD, MIT	rendering[r]
GLM		MIT	rendering[r], RTMT[r], RTMT2[r]
Qt	5	GPL and LGPLv3	RTMT[r], RTMT2[r], Object Ground Truth Annotator[r]	provides graphical user interface (GUI)
MVS-Texturing*		BSD	RTMT2[r], surface[r]	creates textured surfaces for object models

* these libraries will be installed automatically during V4R's build process (either by downloading source code from a given URL or by using code copies inside V4R)

The library itself is independent of any robot operating system such as ROS. To use it inside ROS, you can find wrappers in a separate V4R ROS wrappers repository, which can be placed inside the normal catkin workspace.

Installation:

From Ubuntu Package

First enable the V4R apt repositories (xenial only for now) with the following commands.

• Add the following line to your /etc/apt/sources.list file or in a new file (eg. v4r.list) in the /etc/apt/sources.list.d directory. deb [arch=amd64] https://rwiki.acin.tuwien.ac.at/apt/v4r-release xenial main

• Add the gpg key I used to sign the release to your keyring wget -qO - https://rwiki.acin.tuwien.ac.at/apt/v4r-release/Public.key | sudo apt-key add -

• Update and install sudo apt-get update && sudo apt-get install v4r

From Source

cd ~/somewhere
git clone 'https://rgit.acin.tuwien.ac.at/v4r/v4r.git'
cd v4r
./setup.sh
mkdir build
cd build
cmake ..
make
sudo make install #optional

ROS Interfaces

In order to use V4R in ROS, use the v4r_ros_wrappers.

Notes

Caffe

If you want to use CNN feature extraction, you need to install the Caffe library. We recommend to use CMake-based installation for Caffe and provide the install folder to V4R's cmake call as

cmake .. -DCaffe_DIR=/your_caffe_ws/build/install/share/Caffe