Computers are pretty good at managing 2D images. But at Purdue, researchers are using machine learning to create newly-generated 3D objects from those 2D images — even when only one side is visible. This method could potentially make 3D imagery for augmented reality (AR) and virtual reality (VR) much easier to create. It even allows the computer to «hallucinate» new objects, such as what a car/truck hybrid may look like.
Source: Youtube // Purdue University Mechanical Engineering July 27, 2017
SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks
Current 3D learning paradigms for predictive and generative tasks using convolutional neural networks focus on a voxelized representation of the object.
[…]
Here we study the problem of directly generating the 3D shape surface of rigid and non-rigid shapes using deep convolutional neural networks.
[…]
Our experiments indicate that our network learns a meaningful representation of shape surfaces allowing it to interpolate between shape orientations and poses.
SurfNet: Generating 3D Shape Surfaces Using Deep Residual Networks: Ayan Sinha, Asmi Unmesh, Qixing Huang, Karthik Ramani
Source: https://arxiv.org/abs/1703.04079 // 12 Mar 2017
Purdue University Mechanical Engineering
https://engineering.purdue.edu/ME