UMAD: Unsupervised Mask-Level Anomaly Detection for Autonomous Driving

TL;DR

UMAD introduces an unsupervised approach for anomaly detection in autonomous driving, utilizing generative models and image segmentation to improve detection without relying on labeled outlier data.

Contribution

The paper presents UMAD, a novel unsupervised anomaly detection method that outperforms existing approaches by not requiring labeled outlier data or supervised segmentation models.

Findings

UMAD surpasses state-of-the-art unsupervised anomaly detection methods.

It effectively detects atypical traffic scenarios in autonomous driving.

The approach reduces reliance on labeled outlier data.

Abstract

Dealing with atypical traffic scenarios remains a challenging task in autonomous driving. However, most anomaly detection approaches cannot be trained on raw sensor data but require exposure to outlier data and powerful semantic segmentation models trained in a supervised fashion. This limits the representation of normality to labeled data, which does not scale well. In this work, we revisit unsupervised anomaly detection and present UMAD, leveraging generative world models and unsupervised image segmentation. Our method outperforms state-of-the-art unsupervised anomaly detection.