mmclassification-ros

This is a ROS package for 2D classification, which simply utilizes the toolbox MMClassification of OpenMMLab.

Requirements

• ROS Melodic
• Python 3.6+, PyTorch 1.3+, CUDA 9.2+ and MMCV

To-Do List

• Finish .launch file

  • Check default ros image_topic for raw image (on our own system)
  • decide whether to delete the image_view node: NO!

• Finish srv file

  • figure out the output format (string?): NO!
  • use it!

• Finish script

  • finish line 46 checkpoint

• generate CMakeLists.txt and package.xml

ROS Interfaces

params

• ~publish_rate: the debug image publish rate. default: 1hz
• ~is_service: whether or not to use service instead of subscribe-publish
• ~visualization: whether or not to show the debug image

topics

• ~debug_image: publish the debug image
• ~objects: publish the inference result, containing the classification
• ~image: subscribe the input image. The default one is '/camera/rgb/image_raw' for astra_pro

Example

If you are using the astra_pro camera, you could

source ~/mmcls/bin/activate # in the python3 virtual env
roslaunch mmclassification_ros mmclassification.launch

will continously do the inference on the camera image and publish the detection results