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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Surveillance cameras today often capture NIR (near infrared) images inlow-light environments. However, most face datasets accessible for training andverification are only collected in the VIS (visible light) spectrum. It remainsa challenging problem to match NIR to VIS face images due to the differentlight spectrum. Recently, breakthroughs have been made for VIS face recognitionby applying deep learning on a huge amount of labeled VIS face samples. Thesame deep learning approach cannot be simply applied to NIR face recognitionfor two main reasons: First, much limited NIR face images are available fortraining compared to the VIS spectrum. Second, face galleries to be matched aremostly available only in the VIS spectrum. In this paper, we propose anapproach to extend the deep learning breakthrough for VIS face recognition tothe NIR spectrum, without retraining the underlying deep models that see onlyVIS faces. Our approach consists of two core components, cross-spectralhallucination and low-rank embedding, to optimize respectively input and outputof a VIS deep model for cross-spectral face recognition. Cross-spectralhallucination produces VIS faces from NIR images through a deep learningapproach. Low-rank embedding restores a low-rank structure for faces deepfeatures across both NIR and VIS spectrum. We observe that it is often equallyeffective to perform hallucination to input NIR images or low-rank embedding tooutput deep features for a VIS deep model for cross-spectral recognition. Whenhallucination and low-rank embedding are deployed together, we observesignificant further improvement; we obtain state-of-the-art accuracy on theCASIA NIR-VIS v2.0 benchmark, without the need at all to re-train therecognition system.

Not Afraid of the Dark: NIR-VIS Face Recognition via Cross-spectral Hallucination and Low-rank Embedding