Standardized lossy video coding is at the core of almost all real-world video processing pipelines. Rate control is used to enable standard codecs to adapt to different network bandwidth conditions or storage constraints. However, standard video codecs (e.g., H.264) and their rate control modules aim to minimize video distortion w.r.t. human quality assessment. We demonstrate empirically that standard-coded videos vastly deteriorate the performance of deep vision models. To overcome the deterioration of vision performance, this paper presents the first end-to-end learnable deep video codec control that considers both bandwidth constraints and downstream deep vision performance, while adhering to existing standardization. We demonstrate that our approach better preserves downstream deep vision performance than traditional standard video coding.
@article{arxiv.2308.16215,
title = {Deep Video Codec Control for Vision Models},
author = {Christoph Reich and Biplob Debnath and Deep Patel and Tim Prangemeier and Daniel Cremers and Srimat Chakradhar},
journal= {arXiv preprint arXiv:2308.16215},
year = {2024}
}
Comments
Accepted at CVPR 2024 Workshop on AI for Streaming (AIS)