PIG-Net: Inception based deep learning architecture for 3D point cloud segmentation

Point clouds, being the simple and compact representation of surface geometry 3D objects, have gained increasing popularity with evolution deep learning networks for classification segmentation tasks. Unlike human, teaching machine to analyze segments an object is a challenging task quite essential in various vision applications. In this paper, we address problem labelling point clouds by proposing inception based network architecture called PIG-Net, that effectively characterizes local global geometric details clouds. features are extracted from transformed input points using proposed layers then aligned feature transform. These aggregated average pooling layer obtain features. Finally, feed concatenated convolution segmenting We perform exhaustive experimental analysis PIG-Net on two state-of-the-art datasets, namely, ShapeNet [1] PartNet [2]. evaluate effectiveness our performing ablation study.