Experiments on Anomaly Detection in Autonomous Driving by Forward-Backward Style Transfers

TL;DR

This paper explores a novel style transfer-based approach for anomaly detection in autonomous driving, but the experiments did not yield significant results, highlighting the challenges in this method.

Contribution

The paper proposes using forward-backward style transfer with generative models for anomaly detection in autonomous driving, an approach inspired by image resynthesis techniques.

Findings

Experiments did not produce significant anomaly detection results.

The approach faced challenges in effectively identifying anomalies.

Sharing negative results to inform future research.

Abstract

Great progress has been achieved in the community of autonomous driving in the past few years. As a safety-critical problem, however, anomaly detection is a huge hurdle towards a large-scale deployment of autonomous vehicles in the real world. While many approaches, such as uncertainty estimation or segmentation-based image resynthesis, are extremely promising, there is more to be explored. Especially inspired by works on anomaly detection based on image resynthesis, we propose a novel approach for anomaly detection through style transfer. We leverage generative models to map an image from its original style domain of road traffic to an arbitrary one and back to generate pixelwise anomaly scores. However, our experiments have proven our hypothesis wrong, and we were unable to produce significant results. Nevertheless, we want to share our findings, so that others can learn from our experiments.