3D Shape Induction from 2D Views of Multiple Objects

TL;DR

This paper introduces PrGANs, a novel deep generative model that learns 3D shape distributions from 2D images without 3D or viewpoint annotations, enabling 3D shape generation and viewpoint prediction.

Contribution

The paper presents a new approach called projective GANs that infers 3D shapes from 2D views without supervision, using a projection module within a GAN framework.

Findings

Produces 3D shapes comparable to models trained on 3D data

Learns disentangled representations of geometry and viewpoint

Can generate novel views from a single input image

Abstract

In this paper we investigate the problem of inducing a distribution over three-dimensional structures given two-dimensional views of multiple objects taken from unknown viewpoints. Our approach called "projective generative adversarial networks" (PrGANs) trains a deep generative model of 3D shapes whose projections match the distributions of the input 2D views. The addition of a projection module allows us to infer the underlying 3D shape distribution without using any 3D, viewpoint information, or annotation during the learning phase. We show that our approach produces 3D shapes of comparable quality to GANs trained on 3D data for a number of shape categories including chairs, airplanes, and cars. Experiments also show that the disentangled representation of 2D shapes into geometry and viewpoint leads to a good generative model of 2D shapes. The key advantage is that our model allows us to predict 3D, viewpoint, and generate novel views from an input image in a completely unsupervised manner.