Probing TryOnGAN

TL;DR

This paper reproduces and investigates TryOnGAN, a virtual try-on model, exploring transfer learning, conditioning methods, and latent space properties, revealing insights into its training dynamics and style transfer capabilities.

Contribution

It systematically probes TryOnGAN's components, including transfer learning effects, pose conditioning methods, and latent space properties, some of which are novel analyses.

Findings

Transfer learning aids initial training but loses benefits over time.

Pose conditioning via concatenation outperforms other methods.

Latent space self-disentangles pose and style, enabling style transfer.

Abstract

TryOnGAN is a recent virtual try-on approach, which generates highly realistic images and outperforms most previous approaches. In this article, we reproduce the TryOnGAN implementation and probe it along diverse angles: impact of transfer learning, variants of conditioning image generation with poses and properties of latent space interpolation. Some of these facets have never been explored in literature earlier. We find that transfer helps training initially but gains are lost as models train longer and pose conditioning via concatenation performs better. The latent space self-disentangles the pose and the style features and enables style transfer across poses. Our code and models are available in open source.