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

Face anti-spoofing (FAS) and face forgery detection play vital roles insecuring face biometric systems from presentation attacks (PAs) and viciousdigital manipulation (e.g., deepfakes). Despite promising performance uponlarge-scale data and powerful deep models, the generalization problem ofexisting approaches is still an open issue. Most of recent approaches focus on1) unimodal visual appearance or physiological (i.e., remotephotoplethysmography (rPPG)) cues; and 2) separated feature representation forFAS or face forgery detection. On one side, unimodal appearance and rPPGfeatures are respectively vulnerable to high-fidelity face 3D mask and videoreplay attacks, inspiring us to design reliable multi-modal fusion mechanismsfor generalized face attack detection. On the other side, there are rich commonfeatures across FAS and face forgery detection tasks (e.g., periodic rPPGrhythms and vanilla appearance for bonafides), providing solid evidence todesign a joint FAS and face forgery detection system in a multi-task learningfashion. In this paper, we establish the first joint face spoofing and forgerydetection benchmark using both visual appearance and physiological rPPG cues.To enhance the rPPG periodicity discrimination, we design a two-branchphysiological network using both facial spatio-temporal rPPG signal map and itscontinuous wavelet transformed counterpart as inputs. To mitigate the modalitybias and improve the fusion efficacy, we conduct a weighted batch and layernormalization for both appearance and rPPG features before multi-modal fusion.We find that the generalization capacities of both unimodal (appearance orrPPG) and multi-modal (appearance+rPPG) models can be obviously improved viajoint training on these two tasks. We hope this new benchmark will facilitatethe future research of both FAS and deepfake detection communities.

Benchmarking Joint Face Spoofing and Forgery Detection with Visual and Physiological Cues