Domain Adaptation for Learned Image Compression with Supervised Adapters

arXiv:2404.15591v1 Announce Type: new
Abstract: In Learned Image Compression (LIC), a model is trained at encoding and decoding images sampled from a source domain, often outperforming traditional codecs on natural images; yet its performance may be far from optimal on images sampled from different domains. In this work, we tackle the problem of adapting a pre-trained model to multiple target domains by plugging into the decoder an adapter module for each of them, including the source one. Each adapter improves …

abstract arxiv compression cs.cv decoding domain domain adaptation domains eess.iv encoding image images natural performance type work