Exploring Self-attention for Image Recognition
This work addresses image recognition by proposing self-attention as an alternative to convolution, showing competitive or superior performance, though it is incremental in adapting attention mechanisms from NLP to vision.
The paper explored self-attention variants for image recognition, finding that pairwise self-attention networks matched or outperformed convolutional models, while patchwise models substantially outperformed them, with potential benefits in robustness and generalization.
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.