Emerging Properties in Self-Supervised Vision Transformers
This work addresses the problem of enhancing vision models for researchers and practitioners by demonstrating novel emergent properties in self-supervised ViTs, though it is incremental in adapting existing self-supervised methods to a new architecture.
The paper investigated whether self-supervised learning imparts unique properties to Vision Transformers (ViTs) compared to convolutional networks, finding that self-supervised ViT features enable explicit semantic segmentation and achieve 80.1% top-1 accuracy on ImageNet with ViT-Base in linear evaluation.
In this paper, we question if self-supervised learning provides new properties to Vision Transformer (ViT) that stand out compared to convolutional networks (convnets). Beyond the fact that adapting self-supervised methods to this architecture works particularly well, we make the following observations: first, self-supervised ViT features contain explicit information about the semantic segmentation of an image, which does not emerge as clearly with supervised ViTs, nor with convnets. Second, these features are also excellent k-NN classifiers, reaching 78.3% top-1 on ImageNet with a small ViT. Our study also underlines the importance of momentum encoder, multi-crop training, and the use of small patches with ViTs. We implement our findings into a simple self-supervised method, called DINO, which we interpret as a form of self-distillation with no labels. We show the synergy between DINO and ViTs by achieving 80.1% top-1 on ImageNet in linear evaluation with ViT-Base.