ProDisc-VAD: An Efficient System for Weakly-Supervised Anomaly Detection in Video Surveillance Applications

TL;DR

ProDisc-VAD introduces a lightweight, efficient framework for weakly-supervised video anomaly detection that leverages prototype interaction and contrastive learning to improve discriminability and robustness, achieving high accuracy with minimal parameters.

Contribution

It proposes ProDisc-VAD, a novel framework combining prototype interaction and pseudo-instance discriminative enhancement for improved weakly-supervised video anomaly detection.

Findings

Achieves 97.98% AUC on ShanghaiTech dataset

Achieves 87.12% AUC on UCF-Crime dataset

Uses only 0.4 million parameters, over 800x fewer than recent methods

Abstract

Weakly-supervised video anomaly detection (WS-VAD) using Multiple Instance Learning (MIL) suffers from label ambiguity, hindering discriminative feature learning. We propose ProDisc-VAD, an efficient framework tackling this via two synergistic components. The Prototype Interaction Layer (PIL) provides controlled normality modeling using a small set of learnable prototypes, establishing a robust baseline without being overwhelmed by dominant normal data. The Pseudo-Instance Discriminative Enhancement (PIDE) loss boosts separability by applying targeted contrastive learning exclusively to the most reliable extreme-scoring instances (highest/lowest scores). ProDisc-VAD achieves strong AUCs (97.98% ShanghaiTech, 87.12% UCF-Crime) using only 0.4M parameters, over 800x fewer than recent ViT-based methods like VadCLIP. Code is available at https://github.com/modadundun/ProDisc-VAD.