Anomaly detection using local regions in road images acquired from a hand-held camera

説明

We investigate a method to accurately detect road images with anomalies using the local regions generated from a small number of reference images. There are few datasets of road images with labeled anomalies acquired by hand-held cameras that are large enough to train an accurate detector. We hence evaluated whether an anomaly road image detector using local regions trained on a small dataset of reference images can increase performance. Experimental results show that the use of local regions instead of whole images significantly improves detection performance on a road dataset collected by the local government of Tottori prefecture.

収録刊行物

詳細情報 詳細情報について

問題の指摘

ページトップへ