%0 Conference Proceedings
%T Slimmable Video Codec
%A Zhaocheng Liu
%A Luis Herranz
%A Fei Yang
%A Saiping Zhang
%A Shuai Wan
%A Marta Mrak
%A Marc Gorriz
%B CVPR 2022 Workshop and Challenge on Learned Image Compression (CLIC 2022, 5th Edition)
%D 2022
%F Zhaocheng Liu2022
%O MACO; 601.379; 601.161
%O exported from refbase (http://refbase.cvc.uab.es/show.php?record=3687), last updated on Thu, 20 Apr 2023 10:59:30 +0200
%X Neural video compression has emerged as a novel paradigm combining trainable multilayer neural net-works and machine learning, achieving competitive rate-distortion (RD) performances, but still remaining impractical due to heavy neural architectures, with large memory and computational demands. In addition, models are usually optimized for a single RD tradeoff. Recent slimmable image codecs can dynamically adjust their model capacity to gracefully reduce the memory and computation requirements, without harming RD performance. In this paper we propose a slimmable video codec (SlimVC), by integrating a slimmable temporal entropy model in a slimmable autoencoder. Despite a significantly more complex architecture, we show that slimming remains a powerful mechanism to control rate, memory footprint, computational cost and latency, all being important requirements for practical video compression.
%U https://ieeexplore.ieee.org/document/9857273
%U http://refbase.cvc.uab.es/files/LHY2022.pdf
%U http://dx.doi.org/10.1109/CVPRW56347.2022.00183
%P 1742-1746