#dino #facebook #selfsupervised
Self-Supervised Learning is the final frontier in Representation Learning: Getting useful features without any labels. Facebook AI's new system, DINO, combines advances in Self-Supervised Learning for Computer Vision with the new Vision Transformer (ViT) architecture and achieves impressive results without any labels. Attention maps can be directly interpreted as segmentation maps, and the obtained representations can be used for image retrieval and zero-shot k-nearest neighbor classifiers (KNNs).
Abstract: 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.
Authors: Mathilde Caron, Hugo Touvron, Ishan Misra, Hervé Jégou, Julien Mairal, Piotr Bojanowski, Armand Joulin (Facebook AI)
0:00 - Intro & Overview
6:20 - Vision Transformers
9:20 - Self-Supervised Learning for Images
13:30 - Self-Distillation
15:20 - Building the teacher from the student by moving average
16:45 - DINO Pseudocode
23:10 - Why Cross-Entropy Loss?
28:20 - Experimental Results
33:40 - My Hypothesis why this works
38:45 - Conclusion & Comments