Please use this identifier to cite or link to this item:
For citation please use:
Main Title: GenIcoNet: Generative Icosahedral Mesh Convolutional Network - Dataset
Author(s): Jain, Hardik
Type: Generic Research Data
Abstract: In the past few decades, the computer vision domain has achieved outstanding success in learning 3D shapes for classification, segmentation and image-based reconstruction. However, deep networks are less explored for the generative task of obtaining new 3D shapes from the learned representation. This problem becomes more prominent for 3D shapes represented as surface meshes, mainly because the mesh structure lacks regularity, an essential property for training deep generative networks. In this work, we remedy this problem by proposing a generative icosahedral mesh convolutional network (GenIcoNet) that learns data distribution of surface meshes. Our end-to-end trainable network learns semantic representations using 2D convolutional filters on the regularized icosahedral meshes. During inference, GenIcoNet can be used to generate new geometrically valid shapes directly as surface meshes. Our experiments for interpolation of latent space demonstrate that GenIcoNet is able to outperform networks trained on intermediate surface mesh representations. The variational autoencoder architecture of GenIcoNet learns meaningful representation which is numerically stable w.r.t. small perturbations, allows performing exploration and combination of surface meshes to generate new meaningful shapes, while maintaining the essential property of mesh manifoldness. Our code is available at
Subject(s): ModelNet10 Surface Mesh
Generative Network
Icosahedral Mesh
Issue Date: 15-Nov-2021
Date Available: 26-Nov-2021
Language Code: en
DDC Class: 005 Computer programming, programs, data
TU Affiliation(s): Fak. 4 Elektrotechnik und Informatik » Inst. Technische Informatik und Mikroelektronik » FG Computer Vision & Remote Sensing
Appears in Collections:Technische Universit├Ąt Berlin » Research Data

Files in This Item:


Format: ZIP Archive | Size: 943.28 MB
Format: Markdown | Size: 959 B

Item Export Bar

This item is licensed under a Creative Commons License Creative Commons