Aggregating Global Features into Local Vision Transformer

TL;DR

This paper introduces a global attention module called MOA for local Vision Transformers, improving performance by aggregating spatial global information with fewer parameters, validated on multiple datasets.

Contribution

The paper proposes the multi-resolution overlapped attention (MOA) module to enhance local Vision Transformers by integrating global information effectively.

Findings

MOA significantly improves classification accuracy.

The approach outperforms previous Vision Transformers.

Fewer parameters are needed for comparable or better performance.

Abstract

Local Transformer-based classification models have recently achieved promising results with relatively low computational costs. However, the effect of aggregating spatial global information of local Transformer-based architecture is not clear. This work investigates the outcome of applying a global attention-based module named multi-resolution overlapped attention (MOA) in the local window-based transformer after each stage. The proposed MOA employs slightly larger and overlapped patches in the key to enable neighborhood pixel information transmission, which leads to significant performance gain. In addition, we thoroughly investigate the effect of the dimension of essential architecture components through extensive experiments and discover an optimum architecture design. Extensive experimental results CIFAR-10, CIFAR-100, and ImageNet-1K datasets demonstrate that the proposed approach outperforms previous vision Transformers with a comparatively fewer number of parameters.