Analyzing and Mitigating JPEG Compression Defects in Deep Learning
This study addresses the practical problem of performance degradation in deep learning models due to lossy JPEG compression for consumer applications.
This paper investigates the impact of JPEG compression on deep learning models across various computer vision tasks and datasets, revealing a significant performance drop at high compression rates. The authors propose a novel, unsupervised artifact correction method to mitigate this penalty.
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.