The content introduces a novel approach to video anomaly detection using masked auto-encoders. The proposed model focuses on reconstructing tokens with high motion gradients, introducing synthetic anomalies for training, and employing self-distillation to enhance anomaly detection performance. Extensive experiments demonstrate the model's efficiency and effectiveness on various benchmarks, achieving competitive AUC scores while processing at an impressive speed of 1655 FPS.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Nicolae-Cata... at arxiv.org 03-12-2024
https://arxiv.org/pdf/2306.12041.pdfDeeper Inquiries