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.
@article{arxiv.2406.06370,
title = {UMAD: Unsupervised Mask-Level Anomaly Detection for Autonomous Driving},
author = {Daniel Bogdoll and Noël Ollick and Tim Joseph and Svetlana Pavlitska and J. Marius Zöllner},
journal= {arXiv preprint arXiv:2406.06370},
year = {2024}
}
Comments
Daniel Bogdoll and No\"el Ollick contributed equally. Accepted for publication at BMVC 2024 RROW workshop