Shift-and-Balance Attention

TL;DR

This paper introduces Shift-and-Balance attention, a novel method that improves deep neural network performance by better regulating the contribution of attention mechanisms, outperforming existing methods like Squeeze-and-Excite and Dynamic Convolution.

Contribution

The paper proposes a new attention regulation technique called Shift-and-Balance, which enhances network accuracy by controlling the influence of attention branches more effectively.

Findings

Shift-and-Balance outperforms Squeeze-and-Excite in accuracy when applied in more layers.

It significantly improves network capacity and size.

Achieves comparable or better accuracy than state-of-the-art Dynamic Convolution.

Abstract

Attention is an effective mechanism to improve the deep model capability. Squeeze-and-Excite (SE) introduces a light-weight attention branch to enhance the network's representational power. The attention branch is gated using the Sigmoid function and multiplied by the feature map's trunk branch. It is too sensitive to coordinate and balance the trunk and attention branches' contributions. To control the attention branch's influence, we propose a new attention method, called Shift-and-Balance (SB). Different from Squeeze-and-Excite, the attention branch is regulated by the learned control factor to control the balance, then added into the feature map's trunk branch. Experiments show that Shift-and-Balance attention significantly improves the accuracy compared to Squeeze-and-Excite when applied in more layers, increasing more size and capacity of a network. Moreover, Shift-and-Balance attention achieves better or close accuracy compared to the state-of-art Dynamic Convolution.