Trilateral convolutional neural network for 3D shape reconstruction of objects from a single depth view. Issue 13 (10th October 2019)

• Record Type: Journal Article
• Authors: Rivera, Patricio; Valarezo Añazco, Edwin; Choi, Mun‐Taek; Kim, Tae‐Seong
• Is Part Of: IET image processing. Volume 13:Issue 13(2019)
• Page Start: 2457
• Page End: 2466
• Publication Date: 2019-10-10
• Subjects: image recognition -- object recognition -- medical image processing -- image reconstruction -- belief networks -- neural nets -- solid modelling -- computer vision -- learning (artificial intelligence) -- image representation; reconstruction performance -- fully connected convolutional neural network -- baseline CNN -- autoencoder CNN -- generative adversarial reconstruction network -- Tri‐CNN -- superior reconstruction results -- trilateral convolutional neural network -- shape reconstruction -- single depth view -- three‐dimensional convolutional neural network -- 3D voxel representation -- partial object surface -- single depth image -- dilated convolutions -- convolutional receptive field; Image processing -- Periodicals; 621.36705
• DOI: 10.1049/iet-ipr.2019.0532
• Languages: English
• ISSNs: 1751-9659

Abstract

Abstract : In this study, the authors propose a novel three‐dimensional (3D) convolutional neural network for shape reconstruction via a trilateral convolutional neural network (Tri‐CNN) from a single depth view. The proposed approach produces a 3D voxel representation of an object, derived from a partial object surface in a single depth image. The proposed Tri‐CNN combines three dilated convolutions in 3D to expand the convolutional receptive field more efficiently to learn shape reconstructions. To evaluate the proposed Tri‐CNN in terms of reconstruction performance, the publicly available ShapeNet and Big Data for Grasp Planning data sets are utilised. The reconstruction performance was evaluated against four conventional deep learning approaches: namely, fully connected convolutional neural network, baseline CNN, autoencoder CNN, and a generative adversarial reconstruction network. The proposed experimental results show that Tri‐CNN produces superior reconstruction results in terms of intersection over union values and Brier scores with significantly less number of model parameters and memory.