This paper addresses the problem of anticipating traffic accidents, which aims to forecast potential accidents before they happen. Real-time anticipation is crucial for safe autonomous driving, yet most methods rely on computationally heavy modules like optical flow and intermediate feature extractors, making real-world deployment challenging.
We introduce RARE (Real-time Accident anticipation with Reused Embeddings), a lightweight framework that capitalizes on intermediate features from a single pre-trained object detector. By eliminating additional feature-extraction pipelines, RARE significantly reduces latency. Furthermore, we introduce a novel Attention Score Ranking Loss, which prioritizes higher attention on accident-related objects over non-relevant ones.
RARE demonstrates a 4-8× speedup over existing approaches on the DAD and CCD benchmarks, achieving a latency of 13.6 ms per frame (73.3 FPS) on an RTX 6000. Despite its reduced complexity, it attains state-of-the-art Average Precision and reliably anticipates imminent collisions in real time, highlighting RARE's potential for safety-critical applications.
RARE leverages intermediate features from a single pre-trained object detector, eliminating the need for multiple heavy feature extractors. Our approach combines scene-level temporal context with object-specific embeddings through an attention mechanism.
RARE achieves real-time performance (73.3 FPS) while maintaining the highest accuracy on both DAD and CCD datasets, demonstrating significant speedup over existing methods.
Our Attention Score Ranking Loss ensures that the model consistently focuses on accident-related objects. The visualization shows how RARE identifies high-risk objects early, even when they appear small or emerge mid-sequence.
@misc{song2025realtimetrafficaccidentanticipation,
title={Real-time Traffic Accident Anticipation with Feature Reuse},
author={Inpyo Song and Jangwon Lee},
year={2025},
eprint={2505.17449},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2505.17449},
}