Real-Time 3D Door Detection and Classification on a Low-Power Device

Abstract

In this paper, we propose two methods for door classification with the goal to help and improve robot navigation in indoor spaces and to be used in other areas and applications since it is not limited to door detection as other related works. Our methods work offline, in low-powered computers as the Jetson Nano, in real-time with the ability to differentiate between open, closed and semi-open doors. We use the 3D object classification, PointNet and real-time semantic segmentation algorithms FastFCN and FC-HarDNet. We built a 3D and RGB dataset using a 3D Realsense camera D435 with door images in several indoor environments that we make freely available. Both methods are analysed taking into account their accuracy and the speed of the algorithm in a low powered computer. We conclude that it is possible to have a door classification algorithm running in real-time on a low-power device.