Weakly-supervised Disentangling with Recurrent Transformations for 3D View Synthesis

TL;DR

This paper introduces a recurrent neural network that synthesizes novel 3D views from a single image, effectively disentangling factors like identity and pose for specific object categories.

Contribution

It presents a novel recurrent convolutional encoder-decoder architecture trained end-to-end for 3D view synthesis from a single image, capable of disentangling latent factors without full supervision.

Findings

High-quality view synthesis for faces and chairs.

Effective disentangling of identity and pose factors.

Demonstrated on Multi-PIE and 3D chair datasets.

Abstract

An important problem for both graphics and vision is to synthesize novel views of a 3D object from a single image. This is particularly challenging due to the partial observability inherent in projecting a 3D object onto the image space, and the ill-posedness of inferring object shape and pose. However, we can train a neural network to address the problem if we restrict our attention to specific object categories (in our case faces and chairs) for which we can gather ample training data. In this paper, we propose a novel recurrent convolutional encoder-decoder network that is trained end-to-end on the task of rendering rotated objects starting from a single image. The recurrent structure allows our model to capture long-term dependencies along a sequence of transformations. We demonstrate the quality of its predictions for human faces on the Multi-PIE dataset and for a dataset of 3D chair models, and also show its ability to disentangle latent factors of variation (e.g., identity and pose) without using full supervision.