Lab - Perception and Learning for Robotics

In this course, you will work on small research projects in teams of 2. As course introduction, we will discuss the most relevant papers together in the group.

In the final session of the course, all project outcomes will be presented in a poster session.

Lecture Dates

The lecture will be mixed between virtual and in-person meetings. In-person meetings take place in the computer science building (Friedrich-Hirzebruch-Allee 8).

First Meeting Apr 8, 10.15 - 11.45 room 0.011
Paper Discussion: SpatialBot Apr 15, 10.15 - 11.45 room 0.011
Paper Discussion: Open-Set Recognition in the Age of Vision-Language Models Apr 15, 10.15 - 11.45 room 0.011
Paper Discussion: Understanding Human Hands in Contact at Internet Scale Apr 22, 10.15 - 11.45 zoom link
individual project group meetings
mid-term presentations May 27, 10.15 - 11.45 0.041
Final Poster Presentations Aug 26 room 0.011

Possible Projects

Students can either choose from proposed projects in the first lecture or may propose their own projects if they fit into the course context.

This year, the following projects are already assigned:

  • VLMs with Depth
  • Open-Vocabulary Anomaly Segmentation
  • Exocentric Hand Tracking

As an example for course projects, projects in the last course were:

  • stereo depth completion for meta aria glasses
  • building-scale scene graph creation and language queries from rgbd
  • detection and affordance prediction of functional elements

Course Material

The slides from the initial lecture and any other additional course material are available in this sciebo folder.

Workstations

For this course we provide two Linux workstations in room 0.039. You can go to this lab room and use the workstations locally, or you can login with your CS account at ssh <cs account>@rpllab1.rpl.uni-bonn.de and ssh <cs account>@rpllab5.rpl.uni-bonn.de.

These workstations are shared. You can use them as access terminals, to debug code, download datasets, visualize data, etc. Please make sure to not interfer with each other's data and workspaces and don't change the system libraries. All heavy compute should be done on the cluster.

Questions

For any questions regarding this course, please contact blumh@uni-bonn.de.