Search papers, labs, and topics across Lattice.
4
0
6
Current DeepFake detectors can be fooled by semantically inconsistent real audio and video, highlighting a critical blind spot in their ability to assess realistic manipulations.
Domain shift in object detection is more complex than in classification, demanding nuanced adaptation strategies across multiple pipeline stages.
Forget azimuthal averaging: SRL-MAD learns frequency-aware spectral projections to spot face morphing attacks better than supervised methods, even without attack data.
Forget what you know: explicitly unlearning non-identity-related facial attributes significantly boosts face recognition accuracy, suggesting current models rely on spurious correlations.
