A test-time adaptation method using uncertainty-aware negative learning and gradient masking improves deepfake detector performance under unknown postprocessing and distribution shifts.
Benchmarking neural network robustness to common corruptions and perturbations
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Think Twice before Adaptation: Improving Adaptability of DeepFake Detection via Online Test-Time Adaptation
A test-time adaptation method using uncertainty-aware negative learning and gradient masking improves deepfake detector performance under unknown postprocessing and distribution shifts.