Accident risk rises progressively to improve early warnings in driving
Progressive Risk Estimation for Accident Anticipation
Computer Vision and Pattern Recognition
Summary
Accident anticipation is about spotting signs that a car crash might happen before it actually occurs, but without raising false alarms too early. The authors developed a method called PRE-ACT that treats accident risk as a signal that grows stronger as a crash gets closer in time. This method helps to better predict when an accident might happen, not just whether it will, by watching how risk changes progressively. They also created a way to measure how well these risk predictions behave over time. Their approach showed improvements using real driving datasets.
What this means in practice
- •For automotive safety engineers: Enhance advanced driver assistance systems by predicting how accident risk evolves to better time warnings.
- •For fleet safety managers: Improve monitoring tools to identify increasing accident risk during vehicle operation, enabling earlier interventions.
Authors
Samet Hicsonmez, Eray Çakar, Nermin Samet, Fatma Güney
Abstract
Accident anticipation aims to recognize anomalous driving cues before a crash while avoiding false alarms during normal driving. Existing approaches typically formulate this task as binary classification, focusing on whether an accident will occur rather than when it will occur. We propose PRE-ACT, a framework that models accident risk as a continuously evolving signal that increases as the crash approaches. By explicitly enforcing temporal ordering and distance-to-accident awareness, our method progressively raises risk while suppressing premature alarms, leading to significant improvements on MM-AU subsets and Nexar. We further introduce a Separation Score to evaluate the global behavior of predicted risk curves beyond local temporal windows. Code and visualizations are available at https://github.com/giddyyupp/PRE-ACT.