ROSMLDeploy enables versatile and scalable implementation of processing functionality in ROS2 contexts. It is aimed towards enabling deployment of machine learning models, but developers are free in the processes they want to implement. It includes automatic setup of ROS2 topics, publishers, and subscribers from config files for both processing and recording of data.
You can find the documentation here
There is currently only one way to get started with the library: Use the source version from the repository.
To install the source version from the repository, run the following commands:
git clone https://github.com/WBK-Robotics/ros_ml_deploy.git
cd ros_ml_deploy
pip install src/processing_node
Starting from this, you can import the ProcessingNode into your python project using
from processing_node.processing_node import ProcessingNodeTo use the RecorderNode, you need to additionally run
source install/setup.bash
before running
ros2 run processing_node recorder --out_folder /path/to/output/folder --config_path /path/to/config --num 150
with the following arguments:
-
out_folder: Path to output folder. Mandatory.
-
config_path: Path to config.yaml specifying the relevant Inputs. Mandatory.
-
num: Number of input points to record. Optional, defaults to "-1", i.e. record until manual stoppage.