Learning 3D-aware Image Synthesis with Unknown Pose Distribution

TL;DR

This paper introduces PoF3D, a novel method for 3D-aware image synthesis that eliminates the need for pre-estimated pose priors by jointly learning pose inference and image generation, achieving state-of-the-art results.

Contribution

PoF3D is the first approach to perform high-quality 3D-aware image synthesis without relying on 3D pose priors, using a joint adversarial training framework.

Findings

Achieves comparable image and geometry quality to state-of-the-art methods.

Successfully infers pose distribution automatically from data.

Demonstrates feasibility of pose-free 3D-aware synthesis.

Abstract

Existing methods for 3D-aware image synthesis largely depend on the 3D pose distribution pre-estimated on the training set. An inaccurate estimation may mislead the model into learning faulty geometry. This work proposes PoF3D that frees generative radiance fields from the requirements of 3D pose priors. We first equip the generator with an efficient pose learner, which is able to infer a pose from a latent code, to approximate the underlying true pose distribution automatically. We then assign the discriminator a task to learn pose distribution under the supervision of the generator and to differentiate real and synthesized images with the predicted pose as the condition. The pose-free generator and the pose-aware discriminator are jointly trained in an adversarial manner. Extensive results on a couple of datasets confirm that the performance of our approach, regarding both image quality and geometry quality, is on par with state of the art. To our best knowledge, PoF3D demonstrates the feasibility of learning high-quality 3D-aware image synthesis without using 3D pose priors for the first time.