A leading-silence artifact in FakeAVCeleb and AV-Deepfake1M lets a trivial classifier separate real from fake, and an unsupervised alignment method trained only on real data avoids relying on this shortcut.
Is synthetic voice detection research going into the right direction? InCVPRW, pages 71–80, 2022
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Circumventing shortcuts in audio-visual deepfake detection datasets with unsupervised learning
A leading-silence artifact in FakeAVCeleb and AV-Deepfake1M lets a trivial classifier separate real from fake, and an unsupervised alignment method trained only on real data avoids relying on this shortcut.