Revisiting RCAN: Improved Training for Image Super-Resolution

TL;DR

This paper revisits the RCAN model for image super-resolution, demonstrating that with improved training strategies, RCAN can outperform many newer models, highlighting underfitting as a key issue.

Contribution

The study shows that proper training strategies significantly enhance RCAN's performance, establishing it as a strong baseline for future SR research.

Findings

RCAN can outperform many recent CNN-based SR models with proper training.

Underfitting, not overfitting, limits RCAN's capacity despite its large size.

Increasing training iterations improves performance; regularization may harm results.

Abstract

Image super-resolution (SR) is a fast-moving field with novel architectures attracting the spotlight. However, most SR models were optimized with dated training strategies. In this work, we revisit the popular RCAN model and examine the effect of different training options in SR. Surprisingly (or perhaps as expected), we show that RCAN can outperform or match nearly all the CNN-based SR architectures published after RCAN on standard benchmarks with a proper training strategy and minimal architecture change. Besides, although RCAN is a very large SR architecture with more than four hundred convolutional layers, we draw a notable conclusion that underfitting is still the main problem restricting the model capability instead of overfitting. We observe supportive evidence that increasing training iterations clearly improves the model performance while applying regularization techniques generally degrades the predictions. We denote our simply revised RCAN as RCAN-it and recommend practitioners to use it as baselines for future research. Code is publicly available at https://github.com/zudi-lin/rcan-it.