Analyzing and Mitigating JPEG Compression Defects in Deep Learning

TL;DR

This paper investigates how JPEG compression affects deep learning performance in computer vision tasks and introduces a novel artifact correction method to mitigate these effects without requiring labeled data.

Contribution

It provides a comprehensive analysis of JPEG compression impacts on neural networks and proposes a new artifact correction technique that is label-free and improves robustness.

Findings

High JPEG compression significantly reduces model accuracy.

The proposed artifact correction method effectively mitigates compression artifacts.

Mitigation improves performance on compressed images across multiple datasets.

Abstract

With the proliferation of deep learning methods, many computer vision problems which were considered academic are now viable in the consumer setting. One drawback of consumer applications is lossy compression, which is necessary from an engineering standpoint to efficiently and cheaply store and transmit user images. Despite this, there has been little study of the effect of compression on deep neural networks and benchmark datasets are often losslessly compressed or compressed at high quality. Here we present a unified study of the effects of JPEG compression on a range of common tasks and datasets. We show that there is a significant penalty on common performance metrics for high compression. We test several methods for mitigating this penalty, including a novel method based on artifact correction which requires no labels to train.