[Documentation] [TitleIndex] [WordIndex

Welcome to the second lab session of CoTeSys-ROS Fall School on Cognition-enabled Mobile Manipulation on 2D/3D Perception.

The challenge of the day will be to get 2D/3D Perception using OpenCV and PCL working with both simulated camera/laser data in Gazebo, as well as with data coming from the robots.

The concrete tasks are:

Pre-work

This will guarantee that you will get any changes that we made to the repository.

This might take a few minutes, so please be patient. Let us know if you encounter any errors.

2D Perception: Planar textured object detection

OUR SOLUTION: match_desc.cpp

How the results depend on training/test object, detector and descriptor types, ransac reprojection threshold?

3D Perception: Segmentation and Recognition

rosbag play <bag_file>

What changes? What fails? What parameters do you need to change to make things work?

Advanced task: use 2d inliers together with pcl to find object point cloud in 3D


2024-12-07 12:14