Exploiting the Potential of Standard Convolutional Autoencoders for Image Restoration by Evolutionary Search

TL;DR

This paper demonstrates that simple convolutional autoencoders, optimized via evolutionary search, can outperform complex state-of-the-art image restoration methods that use adversarial training.

Contribution

It introduces an evolutionary algorithm to automatically design effective CAE architectures, achieving superior image restoration performance with standard components.

Findings

Achieved 27.8 dB PSNR on CelebA dataset

Achieved 40.4 dB PSNR on SVHN dataset

Outperformed previous state-of-the-art methods in PSNR

Abstract

Researchers have applied deep neural networks to image restoration tasks, in which they proposed various network architectures, loss functions, and training methods. In particular, adversarial training, which is employed in recent studies, seems to be a key ingredient to success. In this paper, we show that simple convolutional autoencoders (CAEs) built upon only standard network components, i.e., convolutional layers and skip connections, can outperform the state-of-the-art methods which employ adversarial training and sophisticated loss functions. The secret is to employ an evolutionary algorithm to automatically search for good architectures. Training optimized CAEs by minimizing the $\ell_2$ loss between reconstructed images and their ground truths using the ADAM optimizer is all we need. Our experimental results show that this approach achieves 27.8 dB peak signal to noise ratio (PSNR) on the CelebA dataset and 40.4 dB on the SVHN dataset, compared to 22.8 dB and 33.0 dB provided by the former state-of-the-art methods, respectively.