Saliency map based analysis for prediction of car driving difficulty in Google street view scenes
Przemysław Skurowski, Marcin Paszkuta · AIP conference proceedings · 2018
The paper describes how the saliency maps – computational models of visual attention can be used to predict car driving difficulty. The peak number and areas are analyzed for the images are obtained from Google Street View. As a ground truth the crowd sourced average human responses on the speed and subjective driving difficulty are used.