Semantic Segmentation for Street Scene Perception via Mobility-Scooter-Mounted Omnidirectional Camera

カテゴリ: 部門大会

論文No: SS3-3

グループ名: 【C】2022年電気学会電子・情報・システム部門大会

発行日: 2022/08/24

タイトル(英語): Semantic Segmentation for Street Scene Perception via Mobility-Scooter-Mounted Omnidirectional Camera

著者名: Zhang Bowen（甲南大学）,Tanaka Masahiro（甲南大学）

著者名(英語): Bowen Zhang (Konan University),Masahiro Tanaka (Konan University)

キーワード: Semantic Segmentation|Computer Vision|Deep Learning|Panoramic View|Loss Function

要約(日本語): In this work, we assume the road segmentation and obstacles detection in the street scene, and recognize it by semantic segmentation using the mobility-scooter-mounted omnidirectional camera. There is a problem of misrecognition caused by the recognition labels entering the background label. Also, since semantic segmentation can't separate the same object, a 0-1 matrix for masks is created to reduce the error for backpropagation in order to reduce the redundant information obtained there. Finally, we will conduct a ablation study on our created dataset and Cityscapes.