An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

TL;DR

This paper demonstrates that pure transformer models, applied directly to image patches, can outperform traditional convolutional neural networks in image classification tasks when trained on large datasets.

Contribution

It introduces the Vision Transformer (ViT), showing that transformers can be effectively used for image recognition without convolutional components, achieving state-of-the-art results.

Findings

ViT outperforms CNNs on multiple benchmarks.

Pre-trained ViT requires less computational resources.

Large-scale pre-training is crucial for ViT performance.

Abstract

While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.