tech-session-visual-odometry

Contains code and docker deployment files for the Visual Odometry Tech Session that was organized for Kapernikov consultants to familiarize with ROS, visual odometry, visual slam, and robotics software development in general. The presentation for this tech-session can be found on the following link: kapernikov-vo-tech-session

Instructions on how to use this repository and how to do the exercises can be found on the wiki page

Environment Installation
Extras
- OpenCV Installation
- RTAB-Map Installation

Environment Installation

By following this guide, you can create a working environment for experimentation with visual odometry and visual slam. There are three different ways this can be accomplished.

The first option is to install ros, this package and all required libraries natively in an Ubuntu 18.04 OS or Debian Stretch. There is also experimental support for Arch Linux.
The second option is to install a specially prepared docker container that will work in a Linux host OS, but there may be issues with exported graphics in a Windows host OS.
The third and final option is to install a virtual machine image containing an Ubuntu 18.04 OS with all required libraries.

Recommendation:

Native installation for Ubuntu/Debian/Arch Linux users
Docker installation for non-Ubuntu/Debian/Arch Linux users
Virtual machine installation for Windows users

Native-Linux

Supported Operating Systems: Ubuntu, Debian and possibly Arch Linux

Install ROS Melodic Morenia
Install catkin tools

$ sudo apt -y install python-catkin-tools

Initialize the ROS workspace

$ mkdir -p ~/catkin_ws/src
$ cd ~/catkin_ws && catkin init

Clone this repository:

$ git clone https://gitlab.com/Kapernikov/Intern/tech-session-visual-odometry.git ~/catkin_ws/src/tech-session-visual-odometry.git

Install dependencies

$ cd ~/catkin_ws && rosdep install -yr --from-paths src --ignore-src --rosdistro melodic

Build the workspace

$ catkin build

Docker

Warning! If you have nvidia drivers active there issues with some of the graphical applications, so it's recommended to switch to intel drivers if available:

$ sudo prime-select intel

Then reboot and everything should work. If you don't have an nvidia card at all, it should work with the integrated intel card.

You can also try to use nvidia-docker2 instead of docker, although it hasn't been tested.

If this fails you can try one of the two setup approaches (native, VM).

Install docker for Linux
Clone this repository:

$ git clone https://gitlab.com/Kapernikov/Intern/tech-session-visual-odometry.git

Build the image:

$ cd /path/to/tech-session-visual-odometry
$ ./docker_build.sh

Create the container from the built image:

$ ./docker_run.sh

Connect a terminal to the running container

$ ./docker_connect.sh

Virtual Machine

Download and install Virtual-Box 5.2 (NOT 6.0) from the official website
Download the Virtual Machine image. It will also be passed around in USB flash drive on the day of the tech session.
Open Virtual Box 5.2
Go to File > Import Appliance and browse to find the VM image
In the next screen you can change settings such as allocated CPU or RAM. Opt for at least half of your system total. For example if you have 8 cores and 16GB RAM allocate 4 cores and 8GB RAM to the VM.
After finishing the import process, select the VM in the VBox manager and start it
Log in to user: kapernikov with password: kapernikov
Go to ~/catkin_ws/src/tech-session-visual-odometry and pull changes

$ cd ~/catkin_ws/src/tech-session-visual-odometry
$ git pull

Extras

This step is optional and applicable for the more advanced exercises.

Installing OpenCV (3.4.5) with opencv-contrib will include xfeatures2d (and cuda) enabling the use of different features (eg. SIFT, SURF etc) in RTAB-Map. RTAB-Map and RTAB-Map ROS must be installed from source in order to access these features. Furthermore, before building RTAB-Map and RTAB-Map ROS from source, you can install other 3rd party visual odometry and visual slam libraries (eg. viso2, orb2-slam etc) and then RTAB-Map will be able to be built with those libraries and run with different visual odometry or vslam algorithms other than the default F2M and F2F.

OpenCV Installation

Warning: Takes a long time! Prepare in advance!

Install dependencies

$ sudo add-apt-repository "deb http://security.ubuntu.com/ubuntu xenial-security main"
$ sudo apt-get -y install build-essential cmake libgtk2.0-dev pkg-config libavcodec-dev \
  libavformat-dev libswscale-dev python-dev python-numpy libtbb2 libtbb-dev \
  libjpeg-dev libpng-dev libtiff-dev libdc1394-22-dev checkinstall yasm \
  gfortran libjpeg8-dev libjasper-dev software-properties-common libjasper1 \
  libtiff-dev libavcodec-dev libavformat-dev libswscale-dev libdc1394-22-dev \
  libxine2-dev libv4l-dev libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev \
  libgtk2.0-dev libtbb-dev qt5-default libatlas-base-dev libfaac-dev \
  libmp3lame-dev libtheora-dev libvorbis-dev libxvidcore-dev \
  libopencore-amrnb-dev libopencore-amrwb-dev libavresample-dev x264 v4l-utils \
  python3-testresources python3-dev python3-pip python-pip pytho-dev \
  libprotobuf-dev protobuf-compiler libgoogle-glog-dev libgflags-dev \
  libgphoto2-dev libeigen3-dev libhdf5-dev doxygen python-numpy python3-numpy

$ cd /usr/include/linux && sudo ln -s -f ../libv4l1-videodev.h videodev.h && cd -

Clone the repositories:

$ git clone -b 3.4.5 https://github.com/opencv/opencv.git ~/opencv
$ git clone -b 3.4.5 https://github.com/opencv/opencv_contrib.git ~/opencv_contrib
$ git clone https://github.com/ros-perception/vision_opencv.git ~/catkin_ws/src

Build OpenCV without CUDA

$ cd opencv && mkdir build && cd build
$ cmake .. \
    -DCMAKE_BUILD_TYPE=RELEASE \
    -DCMAKE_INSTALL_PREFIX=/usr/local \
    -DOPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
    -DOPENCV_ENABLE_NONFREE=ON \
    -DWITH_QT=ON \
    -DWITH_OPENGL=ON
$ make -j4
$ sudo make install

Build OpenCV with CUDA

$ cd opencv && mkdir build && cd build
$ cmake .. -D CMAKE_BUILD_TYPE=RELEASE \
    -D CMAKE_INSTALL_PREFIX=/usr/local \
    -D WITH_CUDA=ON \
    -D ENABLE_FAST_MATH=1 \
    -D CUDA_FAST_MATH=1 \
    -D WITH_CUBLAS=1 \
    -D OPENCV_EXTRA_MODULES_PATH=../../opencv_contrib/modules \
    -DBUILD_opencv_cudacodec=OFF \
    -DWITH_NVCUVID=OFF \
    -DOPENCV_ENABLE_NONFREE=ON \
    -DWITH_QT=ON \
    -DWITH_OPENGL=ON
$ make -j4
$ sudo make install
$ cd ~/catkin_ws && catkin build vision_opencv

RTAB-Map installation

Warning: Takes a long time! Prepare in advance!

Install dependencies (g2o, GTSAM) by following the instructions in the official installation page
Clone the repositories

$ git clone https://github.com/introlab/rtabmap.git ~/rtabmap
$ git clone https://github.com/introlab/rtabmap_ros.git ~/catkin_ws/src/rtabmap_ros

(Optional) Build another visual odometry or vslam library that integrates with rtabmap such as libfovis, libviso2, dvo_core, okvis, msckf_vio, ORB_SLAM2. For example we will install libviso2:

$ git clone [email protected]:srv/viso2.git ~/viso2
$ cd ~/viso2/libviso2 && mkdir build && cd build
$ cmake ..
$ make -j4
$ sudo make install

Build RTAB-Map

$ cd ~/rtabmap
$ mkdir build && cd build
$ cmake ..
$ make -j4
$ sudo make install

When executing the cmake you will see the libraries that are installed and will be integrated with RTAB-Map.

Build RTAB-Map ROS

$ cd ~/catkin_ws
$ catkin build rtabmap_ros

jeremylebon / tech-session-visual-odometry Goto Github PK