Towards Compact Single Image Super-Resolution via Contrastive Self-distillation

TL;DR

This paper introduces a contrastive self-distillation framework that compresses and accelerates super-resolution CNNs, maintaining high image quality while reducing model size and computational cost for resource-limited devices.

Contribution

It proposes a novel contrastive loss and a channel-splitting network construction method for effective model compression and acceleration in super-resolution tasks.

Findings

Effective compression and acceleration of SR models like EDSR, RCAN, and CARN.

Improved PSNR and SSIM scores with reduced model complexity.

Code availability facilitates reproducibility and further research.

Abstract

Convolutional neural networks (CNNs) are highly successful for super-resolution (SR) but often require sophisticated architectures with heavy memory cost and computational overhead, significantly restricts their practical deployments on resource-limited devices. In this paper, we proposed a novel contrastive self-distillation (CSD) framework to simultaneously compress and accelerate various off-the-shelf SR models. In particular, a channel-splitting super-resolution network can first be constructed from a target teacher network as a compact student network. Then, we propose a novel contrastive loss to improve the quality of SR images and PSNR/SSIM via explicit knowledge transfer. Extensive experiments demonstrate that the proposed CSD scheme effectively compresses and accelerates several standard SR models such as EDSR, RCAN and CARN. Code is available at https://github.com/Booooooooooo/CSD.