Authors:
Roman Janovský
;
David Sedláček
and
Jiří Žára
Affiliation:
Faculty of Electrical Engineering, Czech Technical University in Prague, Technicka 2, Praha 6 and Czechia
Keyword(s):
Point Cloud, Segmentation, Neural Network, U-net, Voxel Grid.
Related
Ontology
Subjects/Areas/Topics:
Applications
;
Informatics in Control, Automation and Robotics
;
Pattern Recognition
;
Robotics
;
Robotics and Automation
;
Software Engineering
;
Virtual Environment, Virtual and Augmented Reality
Abstract:
This paper presents a review of various techniques for improving the performance of neural networks on segmentation task using 3D convolutions and voxel grids – we provide comparison of network with and without max pooling, weighting, masking out the segmentation results, and oversampling results for imbalanced training dataset. We also present changes to 3D U-net architecture that give better results than the standard implementation. Although there are many out-performing architectures using different data input, we show, that although the voxel grids that serve as an input to the 3D U-net, have limits to what they can express, they do not reach their full potential.