Self-Supervised Video Forensics by Audio-Visual Anomaly Detection
Chao Feng, Ziyang Chen, Andrew Owens
University of Michigan, Ann Arbor
CVPR 2023 (Highlight)
✅ Forensics auto regressive model
✅ Audio-visual synchronization model
[ ] Detection File
Visual encoder code is in folder backbone, audio encoder code is in audio_process.py, and audio-visual synchronization transformer code is av_sync_model.py
Audio-visual synchronization model and Forensics autoregressive model checkpoint Google Drive
Audio-visual synchronization model code is based on vit-pytorch
Decoder only autoregressive model is partially based on memory-compressed-attention
Visual encoder is heavily borrowed from action classifiction