Local Anomaly Detection in Videos using Object-Centric Adversarial Learning

TL;DR

This paper introduces an unsupervised, object-centric adversarial learning framework for detecting local anomalies in videos, focusing on object regions to improve accuracy and efficiency.

Contribution

It presents a novel two-stage adversarial approach that models object appearance and gradients for improved local anomaly detection in videos.

Findings

Achieves competitive or superior results on four public benchmarks.

Effectively models object behavior using appearance and gradient correspondence.

Demonstrates the effectiveness of object-centric adversarial learning for anomaly detection.

Abstract

We propose a novel unsupervised approach based on a two-stage object-centric adversarial framework that only needs object regions for detecting frame-level local anomalies in videos. The first stage consists in learning the correspondence between the current appearance and past gradient images of objects in scenes deemed normal, allowing us to either generate the past gradient from current appearance or the reverse. The second stage extracts the partial reconstruction errors between real and generated images (appearance and past gradient) with normal object behaviour, and trains a discriminator in an adversarial fashion. In inference mode, we employ the trained image generators with the adversarially learned binary classifier for outputting region-level anomaly detection scores. We tested our method on four public benchmarks, UMN, UCSD, Avenue and ShanghaiTech and our proposed object-centric adversarial approach yields competitive or even superior results compared to state-of-the-art methods.