Learning Explicit Object-Centric Representations with Vision Transformers

Oscar Vikström, Alexander Ilin

With the recent successful adaptation of transformers to the vision domain, particularly when trained in a self-supervised fashion, it has been shown that vision transformers can l…