6DoF pose estimation with deep learning

Motivation

To improve the absolute accuracy of robot arms computer vision based approaches are applied. The 6DoF pose of a sample mounted at the robot endeffector has to be estimated.

Topics 

  • Deep learning models for 6DoF pose estimation with subpixel accuracy
  • Keypoint detection based on RGB, RGB-D and CAD data
  • 2D-3D and 3D-3D keypoint correspondence learning
  • Unsupervised/self-supervised deep learning methods for automatically adapting to new sample shapes

Skills

  • Python skills are preferable
  • Computer vision skills are preferable
  • Deep learning skills are preferable

Categories:

Research Sector:

Biomechatronics

Type of thesis:

Bachelor Thesis, Master Thesis, Project Thesis

Major:

Informatics, Engineering, Mechatronics, Medical engineering

Major:

Artificial Intelligence and Machine Learning, Medical engineering, Software Engineering and Deployment

Contact: