Domain-specific self-supervised learning

Domain-specific self-supervised learning

Recently, contrastive self-supervised feature learning has almost closed the gap to supervised feature learning in generic image collections [1,2]. In special domain, often there is insufficient labelled data available to conduct supervised pre-training. Here it seems natural to rely on self-supervised techniques. The task in this thesis will be to investigate how contrastive self-supervised techniques can be applied in specific domains.

Potential domains are satellite images, medical images and video recordings of animal behaviour.

[1] T. Chen et al.: A simple framework for contrastive learning of visual representations
[2] X. Chen et al.: Improved baselines with momentum contrastive learning

Requirements

  • Good mathematical understanding (in particular statistics and linear algebra)
  • Python programming
  • Experience in machine learning

Contact

Timo L├╝ddecke

Neural Data Science Group
Institute of Computer Science
University of Goettingen