Few-Shot Domain Adaptation for Learned Image Compression

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Zhang, Tianyu, Zhang, Haotian, Li, Yuqi, Li, Li, Liu, Dong
Formato: Preprint
Publicado: 2024
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916570918289408
author Zhang, Tianyu
Zhang, Haotian
Li, Yuqi
Li, Li
Liu, Dong
author_facet Zhang, Tianyu
Zhang, Haotian
Li, Yuqi
Li, Li
Liu, Dong
contents Learned image compression (LIC) has achieved state-of-the-art rate-distortion performance, deemed promising for next-generation image compression techniques. However, pre-trained LIC models usually suffer from significant performance degradation when applied to out-of-training-domain images, implying their poor generalization capabilities. To tackle this problem, we propose a few-shot domain adaptation method for LIC by integrating plug-and-play adapters into pre-trained models. Drawing inspiration from the analogy between latent channels and frequency components, we examine domain gaps in LIC and observe that out-of-training-domain images disrupt pre-trained channel-wise decomposition. Consequently, we introduce a method for channel-wise re-allocation using convolution-based adapters and low-rank adapters, which are lightweight and compatible to mainstream LIC schemes. Extensive experiments across multiple domains and multiple representative LIC schemes demonstrate that our method significantly enhances pre-trained models, achieving comparable performance to H.266/VVC intra coding with merely 25 target-domain samples. Additionally, our method matches the performance of full-model finetune while transmitting fewer than $2\%$ of the parameters.
format Preprint
id arxiv_https___arxiv_org_abs_2409_11111
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Few-Shot Domain Adaptation for Learned Image Compression
Zhang, Tianyu
Zhang, Haotian
Li, Yuqi
Li, Li
Liu, Dong
Image and Video Processing
Computer Vision and Pattern Recognition
Learned image compression (LIC) has achieved state-of-the-art rate-distortion performance, deemed promising for next-generation image compression techniques. However, pre-trained LIC models usually suffer from significant performance degradation when applied to out-of-training-domain images, implying their poor generalization capabilities. To tackle this problem, we propose a few-shot domain adaptation method for LIC by integrating plug-and-play adapters into pre-trained models. Drawing inspiration from the analogy between latent channels and frequency components, we examine domain gaps in LIC and observe that out-of-training-domain images disrupt pre-trained channel-wise decomposition. Consequently, we introduce a method for channel-wise re-allocation using convolution-based adapters and low-rank adapters, which are lightweight and compatible to mainstream LIC schemes. Extensive experiments across multiple domains and multiple representative LIC schemes demonstrate that our method significantly enhances pre-trained models, achieving comparable performance to H.266/VVC intra coding with merely 25 target-domain samples. Additionally, our method matches the performance of full-model finetune while transmitting fewer than $2\%$ of the parameters.
title Few-Shot Domain Adaptation for Learned Image Compression
topic Image and Video Processing
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2409.11111