If you haven't used ROS before you will need to go through tutorial 1 of http://wiki.ros.org/ROS/Tutorials to create a catkin workspace.
Copy the retinanet_inferencer folder into your catkin workspace and run catkin_make after installing the requirements. Then open run_inferencer.launch and edit the model_path to point to your .h5 model file which has been converted to inference mode. You can also change the input and output topic paths.
You can then run roslaunch retinanet_inferencer run_inferencer.launch which will bring up the node. You need to feed Image type messages to the /object_detection_input topic (rosrun topic_tools relay /your/camera/feed /object_detection_input) and there will be visual results on /object_detection_output_image which you can view with rqt_image_view. There are also Detection2DArray messages on the /object_detection_output topic.
- Sensor messages (eg sudo apt-get install ros-melodic-sensor-msgs)
- Geometry messages
- Vision messages
- CV Bridge
- OpenCV2
- Keras Retinanet Installed