Conditional Image Synthesis With Auxiliary Classifier GANs

Augustus Odena, Christopher Olah, Jonathon Shlens (privately revealed to you)

TL;DR: New GAN architecture that generates samples from all 1000 ImageNet classes. Two new methods for measuring sample quality and diversity.

Abstract:

Synthesizing high resolution photorealistic images has been a long-standing challenge
in machine learning. In this paper we introduce new methods for the improved
training of generative adversarial networks (GANs) for image synthesis.
We construct a variant of GANs employing label conditioning that results in
128 × 128 resolution image samples exhibiting global coherence. We expand
on previous work for image quality assessment to provide two new analyses for
assessing the discriminability and diversity of samples from class-conditional image
synthesis models. These analyses demonstrate that high resolution samples
provide class information not present in low resolution samples. Across 1000
ImageNet classes, 128 × 128 samples are more than twice as discriminable as artificially
resized 32 × 32 samples. In addition, 84.7% of the classes have samples
exhibiting diversity comparable to real ImageNet data.

Conflicts: google.com

Community Implementations: [ 31 code implementations](https://www.catalyzex.com/paper/arxiv:1610.09585/code)