Compiling cv_bridge for ROS Melodic with Python 3 (System and Conda Environments)

Environment

Ubuntu 18.04, Python 3.6.9 / Conda Python 3.6, ROS Melodic.

1. Compile cv_bridge in native Python 3.6 environment

<code>sudo apt-get install python3-dev python3-numpy python3-yaml ros-melodic-cv-bridge python3-rospkg-modules</code>
<code>pip3 install pip --update</code>
<code>pip3 install rosdep rosinstall catkin_pkg</code>

Initialize workspace:

<code>mkdir -p cv_bridge_ws/src && cd cv_bridge_ws/src</code>
<code>catkin_init_workspace</code>

Clone source:

<code>git clone https://github.com/ros-perception/vision_opencv.git</code>

Compile:

<code>cd ../</code>
<code>catkin_make install -DPYTHON_EXECUTABLE=/usr/bin/python3</code>

Compilation error resolution

Problem 1: Incorrect Boost Python library path. Locate the correct libboost_python* file (e.g., in /usr/lib/aarch64-linux-gnu/ ) and adjust CMakeLists.txt lines 11 and 14 from python37 to python3 , then recompile.

2. Compile cv_bridge in Conda Python 3.x environment

Create Conda environment:

<code>conda create -n ros python=3.6</code>
<code>conda activate ros</code>
<code>pip install --upgrade pip</code>
<code>pip install rosdep rosinstall catkin_pkg rospkg numpy pyyaml opencv-python</code>

Update ~/.bashrc to use conda activate ros and source the workspace after building.

Initialize workspace and clone source:

<code>mkdir -p ros_cv_bridge/src && cd ros_cv_bridge/src</code>
<code>catkin_init_workspace</code>
<code>git clone https://gitee.com/irvingao/vision_opencv.git</code>

Compile with Conda Python interpreter:

<code>cd ../</code>
<code>export CPLUS_INCLUDE_PATH=/home/innox/anaconda3/include/python3.7m</code>
<code>catkin_make install -DCMAKE_BUILD_TYPE=Release -DSETUPTOOLS_DEB_LAYOUT=OFF -DPYTHON_EXECUTABLE=/home/innox/anaconda3/bin/python</code>

Add the generated files to the environment:

<code>vim ~/.bashrc</code>
<code>source ~/ros_cv_bridge/install/setup.bash --extend</code>
<code>source ~/.bashrc</code>

Testing

Open a new terminal and run:

<code>python</code>
<code>import cv_bridge</code>
<code>from cv_bridge.boost.cv_bridge_boost import getCvType</code>

If the imports succeed, cv_bridge has been successfully compiled and can be used in ROS Python 3 nodes.