ResNeSt: Split-Attention Networks

TL;DR

ResNeSt introduces a split-attention architecture that enhances feature interaction and diversity, leading to improved accuracy and transfer learning performance in image recognition tasks.

Contribution

It proposes a modular split-attention block integrated into ResNeSt, which outperforms previous models like EfficientNet and is effective as a backbone in various benchmarks.

Findings

ResNeSt surpasses EfficientNet in accuracy and latency.

Achieves state-of-the-art transfer learning results.

Used in winning entries of COCO-LVIS challenge.

Abstract

It is well known that featuremap attention and multi-path representation are important for visual recognition. In this paper, we present a modularized architecture, which applies the channel-wise attention on different network branches to leverage their success in capturing cross-feature interactions and learning diverse representations. Our design results in a simple and unified computation block, which can be parameterized using only a few variables. Our model, named ResNeSt, outperforms EfficientNet in accuracy and latency trade-off on image classification. In addition, ResNeSt has achieved superior transfer learning results on several public benchmarks serving as the backbone, and has been adopted by the winning entries of COCO-LVIS challenge. The source code for complete system and pretrained models are publicly available.