Design What You Desire: Icon Generation from Orthogonal Application and Theme Labels

TL;DR

This paper introduces IconGAN, a novel generative model that creates customizable icons based on orthogonal application and theme labels, addressing label entanglement issues in icon generation.

Contribution

The paper proposes IconGAN with dual discriminators and a contrastive disentanglement strategy to improve icon generation quality and label disentanglement.

Findings

IconGAN outperforms baseline methods on the AppIcon benchmark.

Disentangling app and theme representations improves icon customization.

Orthogonal augmentation enhances model robustness.

Abstract

Generative adversarial networks (GANs) have been trained to be professional artists able to create stunning artworks such as face generation and image style transfer. In this paper, we focus on a realistic business scenario: automated generation of customizable icons given desired mobile applications and theme styles. We first introduce a theme-application icon dataset, termed AppIcon, where each icon has two orthogonal theme and app labels. By investigating a strong baseline StyleGAN2, we observe mode collapse caused by the entanglement of the orthogonal labels. To solve this challenge, we propose IconGAN composed of a conditional generator and dual discriminators with orthogonal augmentations, and a contrastive feature disentanglement strategy is further designed to regularize the feature space of the two discriminators. Compared with other approaches, IconGAN indicates a superior advantage on the AppIcon benchmark. Further analysis also justifies the effectiveness of disentangling app and theme representations. Our project will be released at: https://github.com/architect-road/IconGAN.