Cross-Layer Retrospective Retrieving via Layer Attention

TL;DR

This paper introduces a cross-layer attention mechanism called MRLA that enhances neural network representations by retrieving information from previous layers, improving performance across vision tasks with minimal additional computation.

Contribution

The paper proposes a novel multi-head recurrent layer attention mechanism that effectively integrates cross-layer information, boosting vision network performance with a lightweight design.

Findings

Improves ResNet-50 Top-1 accuracy by 1.6%.

Enhances dense prediction tasks with 3-4% gains in box and mask AP.

Reduces computational cost with a lightweight version of MRLA.

Abstract

More and more evidence has shown that strengthening layer interactions can enhance the representation power of a deep neural network, while self-attention excels at learning interdependencies by retrieving query-activated information. Motivated by this, we devise a cross-layer attention mechanism, called multi-head recurrent layer attention (MRLA), that sends a query representation of the current layer to all previous layers to retrieve query-related information from different levels of receptive fields. A light-weighted version of MRLA is also proposed to reduce the quadratic computation cost. The proposed layer attention mechanism can enrich the representation power of many state-of-the-art vision networks, including CNNs and vision transformers. Its effectiveness has been extensively evaluated in image classification, object detection and instance segmentation tasks, where improvements can be consistently observed. For example, our MRLA can improve 1.6% Top-1 accuracy on ResNet-50, while only introducing 0.16M parameters and 0.07B FLOPs. Surprisingly, it can boost the performances by a large margin of 3-4% box AP and mask AP in dense prediction tasks. Our code is available at https://github.com/joyfang1106/MRLA.