P‐202: Late‐News Poster: Dynamic Obstacle Detection to Improve BVI Pedestrian's Navigation Decision using CNNs

Abstract

Two CNNs models are proposed to detect dynamic obstacles in urban setting so as to improve BVI pedestrian's navigation decision. First, we re-implement the inspired model, yolov2, on KITTI datasets. Regarding evaluation, the two models perform better than yolov2 from 6.26% to 10.99% on car detection at different difficulty levels.