Conditional Image Synthesis With Auxiliary Classifier GANs
This work addresses the problem of generating realistic conditional images for computer vision applications, representing an incremental improvement over previous GAN methods.
The paper tackles the challenge of synthesizing high-resolution photorealistic images by introducing improved training methods for generative adversarial networks (GANs) with label conditioning, achieving 128x128 resolution samples with global coherence. The results show that 128x128 samples are over twice as discriminable as resized 32x32 samples across 1000 ImageNet classes, and 84.7% of classes exhibit diversity comparable to real data.
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 128x128 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, 128x128 samples are more than twice as discriminable as artificially resized 32x32 samples. In addition, 84.7% of the classes have samples exhibiting diversity comparable to real ImageNet data.