AUTO3D: Novel view synthesis through unsupervisely learned variational viewpoint and global 3D representation

TL;DR

AUTO3D introduces an unsupervised variational framework for novel view synthesis from limited images, disentangling pose and 3D shape without explicit 3D reconstruction, achieving competitive results.

Contribution

It presents a novel unsupervised, end-to-end trainable model that learns 3D representations and poses for view synthesis without supervision.

Findings

Achieves comparable or better results than supervised methods.

Effectively disentangles pose and shape in an unsupervised manner.

Works with limited input views for novel view synthesis.

Abstract

This paper targets on learning-based novel view synthesis from a single or limited 2D images without the pose supervision. In the viewer-centered coordinates, we construct an end-to-end trainable conditional variational framework to disentangle the unsupervisely learned relative-pose/rotation and implicit global 3D representation (shape, texture and the origin of viewer-centered coordinates, etc.). The global appearance of the 3D object is given by several appearance-describing images taken from any number of viewpoints. Our spatial correlation module extracts a global 3D representation from the appearance-describing images in a permutation invariant manner. Our system can achieve implicitly 3D understanding without explicitly 3D reconstruction. With an unsupervisely learned viewer-centered relative-pose/rotation code, the decoder can hallucinate the novel view continuously by sampling the relative-pose in a prior distribution. In various applications, we demonstrate that our model can achieve comparable or even better results than pose/3D model-supervised learning-based novel view synthesis (NVS) methods with any number of input views.