Deep learning model also predicts future accidents

Deep learning model also predicts future accidents

21. October 2021 0 By Horst Buchwald

Deep learning model also predicts future accidents

Berlin, Oct. 21, 2021

A deep-learning model can create high-resolution maps showing areas at risk for car accidents and help predict future accidents.

The model, developed by MIT CSAIL and the Qatar Center for Artificial Intelligence, was able to identify areas that it deemed high-risk, even if no accidents were recorded there. Accidents later occurred in those areas.

The software identifies high-risk locations using satellite imagery showing road infrastructure and GPS trajectory patterns over traffic. It can “describe the expected number of accidents over a period of time in the future.”

The model can also generalize between cities and predict accident maps in unknown areas.

The ability to derive accident maps even without historical data could help find safer routes, design safer roads, and help insurance companies create customized plans based on a driver’s routes.

 

The paper is available here:

https://openaccess.thecvf.com/content/ICCV2021/papers/He_Inferring_High-Resolution_Traffic_Accident_Risk_Maps_Based_on_Satellite_Imagery_ICCV_2021_paper.pdf

Hits: 2