English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools annual

Global: Zendy Free Plan

Global: Zendy Tools monthly

Global: Zendy Plus monthly

The latest advancements in neural image compression show great potential insurpassing the rate-distortion performance of conventional standard codecs.Nevertheless, there exists an indelible domain gap between the datasetsutilized for training (i.e., natural images) and those utilized for inference(e.g., artistic images). Our proposal involves a low-rank adaptation approachaimed at addressing the rate-distortion drop observed in out-of-domaindatasets. Specifically, we perform low-rank matrix decomposition to updatecertain adaptation parameters of the client's decoder. These updatedparameters, along with image latents, are encoded into a bitstream andtransmitted to the decoder in practical scenarios. Due to the low-rankconstraint imposed on the adaptation parameters, the resulting bit rateoverhead is small. Furthermore, the bit rate allocation of low-rank adaptationis \emph{non-trivial}, considering the diverse inputs require varyingadaptation bitstreams. We thus introduce a dynamic gating network on top of thelow-rank adaptation method, in order to decide which decoder layer shouldemploy adaptation. The dynamic adaptation network is optimized end-to-end usingrate-distortion loss. Our proposed method exhibits universality across diverseimage datasets. Extensive results demonstrate that this paradigm significantlymitigates the domain gap, surpassing non-adaptive methods with an averageBD-rate improvement of approximately $19\%$ across out-of-domain images.Furthermore, it outperforms the most advanced instance adaptive methods byroughly $5\%$ BD-rate. Ablation studies confirm our method's ability touniversally enhance various image compression architectures.

Dynamic Low-Rank Instance Adaptation for Universal Neural Image Compression