A Perspective on Deep Vision Performance with Standard Image and Video Codecs
It highlights a critical problem for deploying deep vision on resource-constrained devices that rely on cloud inference, showing incremental analysis by extending to more tasks.
This paper investigates the impact of using standard image and video codecs like JPEG and H.264 on deep vision models, finding that strong compression significantly reduces accuracy, such as an over 80% drop in mIoU for semantic segmentation.
Resource-constrained hardware, such as edge devices or cell phones, often rely on cloud servers to provide the required computational resources for inference in deep vision models. However, transferring image and video data from an edge or mobile device to a cloud server requires coding to deal with network constraints. The use of standardized codecs, such as JPEG or H.264, is prevalent and required to ensure interoperability. This paper aims to examine the implications of employing standardized codecs within deep vision pipelines. We find that using JPEG and H.264 coding significantly deteriorates the accuracy across a broad range of vision tasks and models. For instance, strong compression rates reduce semantic segmentation accuracy by more than 80% in mIoU. In contrast to previous findings, our analysis extends beyond image and action classification to localization and dense prediction tasks, thus providing a more comprehensive perspective.