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

Convolutional Neural Networks (CNNs) have achieved impressive results acrossmany super-resolution (SR) and image restoration tasks. While many suchnetworks can upscale low-resolution (LR) images using just the raw pixel-levelinformation, the ill-posed nature of SR can make it difficult to accuratelysuper-resolve an image which has undergone multiple different degradations.Additional information (metadata) describing the degradation process (such asthe blur kernel applied, compression level, etc.) can guide networks tosuper-resolve LR images with higher fidelity to the original source. Previousattempts at informing SR networks with degradation parameters have indeed beenable to improve performance in a number of scenarios. However, due to thefully-convolutional nature of many SR networks, most of these metadata fusionmethods either require a complete architectural change, or necessitate theaddition of significant extra complexity. Thus, these approaches are difficultto introduce into arbitrary SR networks without considerable designalterations. In this paper, we introduce meta-attention, a simple mechanismwhich allows any SR CNN to exploit the information available in relevantdegradation parameters. The mechanism functions by translating the metadatainto a channel attention vector, which in turn selectively modulates thenetwork's feature maps. Incorporating meta-attention into SR networks isstraightforward, as it requires no specific type of architecture to functioncorrectly. Extensive testing has shown that meta-attention can consistentlyimprove the pixel-level accuracy of state-of-the-art (SOTA) networks whenprovided with relevant degradation metadata. For PSNR, the gain onblurred/downsampled (X4) images is of 0.2969 dB (on average) and 0.3320 dB forSOTA general and face SR models, respectively.

Improving Super-Resolution Performance Using Meta-Attention Layers