Rotation Invariant Graph Neural Network for 3D Point Clouds

In this paper we propose a novel rotation normalization technique for point cloud processing using an oriented bounding box. We use method to create annotation tool part segmentation on real camera data. Custom data sets are used train our network classification and tasks. Successful deployment is completed embedded device with limited power. A comparison made other rotation-invariant features in noisy synthetic datasets. Our offers more auxiliary information related the dimension, position, orientation of object than previous methods while performing at similar level.