BYU

Abstract by Michael Brodie

Personal Infomation


Presenter's Name

Michael Brodie

Degree Level

Doctorate

Abstract Infomation


Department

Computer Science

Faculty Advisor

Tony Martinez

Title

Finetuning CoachGAN

Abstract

The recent CoachGAN algorithm provides an efficient, inference-time output enhancement method for generative models. We introduce a new model health metric called the Discriminator Conditioning Number (DCN). DCN extracts the maximum and minimum eigenvalues of the squared Jacobian of the discriminator for a random input sample. The quotient of these values serves as a compact but informative number that indicates the stability of CoachGAN if used with the current discriminator. Using state-of-the-art GAN models and a variety of datasets, we use FID and IS metrics to quantitatively validate that DCN is an accurate measure of  CoachGAN effectiveness ahead of time. We also introduce a fine-tuning algorithm that improves the DCN and leads to higher FID and IS scores. We demonstrate these improvements across a variety of tasks, including image superresolution, image inpainting, image-to-image translation, conditional image generation, and unconditional image generation.