Bucharest Deep Learning Meetup – Avoiding Shortcuts in Audio-Visual Deepfake Detection Datasets

Presentation by Ștefan Smeu on deepfake detection, the tendency of AI models to learn shortcuts from datasets, and the use of audio-video synchronization for robust detection on new examples. The topic is based on a paper accepted at CVPR 2025.

Details

Organizer
Bucharest Deep Learning Meetup; guest speaker: Ștefan Smeu.
Activity Schedule
March 26, 2026, at 6:30 p.m.
Location and Format
The new IMF building, Politehnica Business Tower, 15G Iuliu Maniu Street, 5th floor, Room 503.
Participation
Limited seating.
Tags :