Recognition of Abnormal Events in Surveillance Videos using Weakly Supervised Dual-Encoder Models

TL;DR

This paper introduces a dual-encoder framework that combines convolutional and transformer models with weak supervision to effectively detect anomalies in surveillance videos, achieving high accuracy on a challenging dataset.

Contribution

The paper proposes a novel weakly supervised dual-encoder model that integrates convolutional and transformer features for anomaly detection in videos.

Findings

Achieved 90.7% AUC on UCF-Crime dataset

Effectively detects rare and diverse anomalies with weak supervision

Combines convolutional and transformer representations for improved performance

Abstract

We address the challenge of detecting rare and diverse anomalies in surveillance videos using only video-level supervision. Our dual-backbone framework combines convolutional and transformer representations through top-k pooling, achieving 90.7% area under the curve (AUC) on the UCF-Crime dataset.