CNN detectors for face-swap videos achieve near-perfect accuracy on the same dataset they are trained on, but cross-dataset accuracy drops dramatically, showing they learn dataset-specific cues rather than occlusion-based artifacts.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Spotting tell-tale visual artifacts in face swapping videos: strengths and pitfalls of CNN detectors
CNN detectors for face-swap videos achieve near-perfect accuracy on the same dataset they are trained on, but cross-dataset accuracy drops dramatically, showing they learn dataset-specific cues rather than occlusion-based artifacts.