Anomaly Detection

Anomaly Detection

Anomaly detection in driving scenes is an important problem in advanced driver assistance systems (ADAS). It is important to identify potential hazard scenarios as early as possible to avoid potential accidents. We consider unsupervised methods to quantify driving anomalies using a conditional generative adversarial network (GAN). The system learns how the driver behaves under different driving condition. The anomalies are detected when the user behavior deviates from the normal behavior predicted by our model. We have trained a driving anomaly detection system from 150 hours of naturalistic driving data and demonstrated that our approach can better identify anomalous situations compared to baseline anomaly detection systems.