Course projects

Olaf Vision

Guiding robots with vision-language models and imitation learning.

  • VLMs
  • Imitation learning
  • React

Course project for Robot Perception and Learning (A+). We combined VLMs and imitation learning so people can guide a robot intuitively.

I built a React web app to collect human-corrected robot trajectories as training data. Along the way we found that VLMs struggle to understand 3D, and we found limits to what imitation learning could do here. Slides.