Predict and Resist: Long-Term Accident Anticipation under Sensor Noise
Xingcheng Liu, Bin Rao, Yanchen Guan, Chengyue Wang, Haicheng Liao, Jiaxun Zhang, Chengyu Lin, Meixin Zhu, Zhenning Li · 2025
Read Predict and Resist: Long-Term Accident Anticipation under Sensor Noise by Xingcheng Liu, Bin Rao, Yanchen Guan, Chengyue Wang, Haicheng Liao, Jiaxun Zhang, Chengyu Lin, Meixin Zhu, Zhenning Li (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-08640.txt.
Description
Xingcheng Liu and eight co-authors build a driving safety system that predicts accidents early even when camera sensors are noisy. Rain, motion blur, and weak hardware can hide the very cues a warning system needs, and alerts must come early without crying wolf.
A diffusion module cleans image and object features step by step, restoring motion and interaction cues lost to noise. In parallel, an actor-critic model weighs long stretches of video and uses time-weighted rewards to pick the best alert moment. On three benchmark datasets the system reaches top accuracy and longer average warning times before impact. It stays robust under Gaussian and impulse noise, and its warnings arrive earlier and steadier in both routine and tangled traffic scenes.
Catalog record
| CREATOR | Xingcheng Liu, Bin Rao, Yanchen Guan, Chengyue Wang, Haicheng Liao, Jiaxun Zhang, Chengyu Lin, Meixin Zhu, Zhenning Li |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2511-08640