Enhanced Anime Image Generation Using USE-CMHSA-GAN

  • 2024-11-17 21:25:24
  • J. Lu
  • 0

Abstract

With the growing popularity of ACG (Anime, Comics, and Games) culture,generating high-quality anime character images has become an important researchtopic. This paper introduces a novel Generative Adversarial Network model,USE-CMHSA-GAN, designed to produce high-quality anime character images. Themodel builds upon the traditional DCGAN framework, incorporating USE and CMHSAmodules to enhance feature extraction capabilities for anime character images.Experiments were conducted on the anime-face-dataset, and the resultsdemonstrate that USE-CMHSA-GAN outperforms other benchmark models, includingDCGAN, VAE-GAN, and WGAN, in terms of FID and IS scores, indicating superiorimage quality. These findings suggest that USE-CMHSA-GAN is highly effectivefor anime character image generation and provides new insights for furtherimproving the quality of generative models.

 

Quick Read (beta)

loading the full paper ...