Deep neural networks have reached human-level performance on many comput...
Today's computer vision models achieve human or near-human level perform...
It is common practice to reuse models initially trained on different dat...
This paper presents Contrastive Reconstruction, ConRec - a self-supervis...