Exploring Self-attention for Image Recognition

TL;DR

This paper investigates different self-attention mechanisms for image recognition, demonstrating that self-attention can outperform traditional convolutional models and may enhance robustness and generalization.

Contribution

It introduces and evaluates pairwise and patchwise self-attention models, showing their effectiveness and advantages over convolutional networks in image recognition tasks.

Findings

Pairwise self-attention matches or exceeds convolutional models.

Patchwise self-attention significantly outperforms convolutional baselines.

Self-attention networks show potential for improved robustness and generalization.

Abstract

Recent work has shown that self-attention can serve as a basic building block for image recognition models. We explore variations of self-attention and assess their effectiveness for image recognition. We consider two forms of self-attention. One is pairwise self-attention, which generalizes standard dot-product attention and is fundamentally a set operator. The other is patchwise self-attention, which is strictly more powerful than convolution. Our pairwise self-attention networks match or outperform their convolutional counterparts, and the patchwise models substantially outperform the convolutional baselines. We also conduct experiments that probe the robustness of learned representations and conclude that self-attention networks may have significant benefits in terms of robustness and generalization.