A Comparative Study of NAFNet Baselines for Image Restoration

TL;DR

This paper evaluates NAFNet, a simple and efficient image restoration model, analyzing how its core components affect performance through ablation studies on noisy and blurred CIFAR10 images.

Contribution

It provides a detailed ablation study of NAFNet's components, validating the effectiveness of SimpleGate, simplified attention, and LayerNorm in image restoration.

Findings

SimpleGate and simplified attention improve restoration results

LayerNorm is crucial for stable training

Component modifications significantly impact PSNR and SSIM

Abstract

We study NAFNet (Nonlinear Activation Free Network), a simple and efficient deep learning baseline for image restoration. By using CIFAR10 images corrupted with noise and blur, we conduct an ablation study of NAFNet's core components. Our baseline model implements SimpleGate activation, Simplified Channel Activation (SCA), and LayerNormalization. We compare this baseline to different variants that replace or remove components. Quantitative results (PSNR, SSIM) and examples illustrate how each modification affects restoration performance. Our findings support the NAFNet design: the SimpleGate and simplified attention mechanisms yield better results than conventional activations and attention, while LayerNorm proves to be important for stable training. We conclude with recommendations for model design, discuss potential improvements, and future work.