Transformer for Single Image Super-Resolution

TL;DR

This paper introduces ESRT, a hybrid Transformer-CNN model for single image super-resolution that achieves competitive results with significantly reduced computational costs and GPU memory usage.

Contribution

The paper proposes a novel Efficient Super-Resolution Transformer (ESRT) combining lightweight CNN and Transformer backbones with an efficient attention mechanism.

Findings

ESRT achieves competitive super-resolution results.

ESRT uses only 4,191M GPU memory compared to 16,057M of original Transformer.

Extensive experiments validate the efficiency and effectiveness of ESRT.

Abstract

Single image super-resolution (SISR) has witnessed great strides with the development of deep learning. However, most existing studies focus on building more complex networks with a massive number of layers. Recently, more and more researchers start to explore the application of Transformer in computer vision tasks. However, the heavy computational cost and high GPU memory occupation of the vision Transformer cannot be ignored. In this paper, we propose a novel Efficient Super-Resolution Transformer (ESRT) for SISR. ESRT is a hybrid model, which consists of a Lightweight CNN Backbone (LCB) and a Lightweight Transformer Backbone (LTB). Among them, LCB can dynamically adjust the size of the feature map to extract deep features with a low computational cost. LTB is composed of a series of Efficient Transformers (ET), which occupies a small GPU memory occupation, thanks to the specially designed Efficient Multi-Head Attention (EMHA). Extensive experiments show that ESRT achieves competitive results with low computational costs. Compared with the original Transformer which occupies 16,057M GPU memory, ESRT only occupies 4,191M GPU memory. All codes are available at https://github.com/luissen/ESRT.