IGAN: A New Inception-based Model for Stable and High-Fidelity Image Synthesis Using Generative Adversarial Networks

TL;DR

The paper introduces IGAN, a novel GAN architecture that employs inception-inspired and dilated convolutions to produce high-quality images with improved training stability, reducing mode collapse and gradient issues.

Contribution

IGAN is the first model to integrate inception-inspired and dilated convolutions for stable, high-fidelity image synthesis in GANs.

Findings

Achieved FID of 13.12 on CUB-200 and 15.08 on ImageNet datasets.

Attained Inception Scores of 9.27 and 68.25, indicating high image diversity.

Demonstrated reduced mode collapse and stabilized training with dropout and spectral normalization.

Abstract

Generative Adversarial Networks (GANs) face a significant challenge of striking an optimal balance between high-quality image generation and training stability. Recent techniques, such as DCGAN, BigGAN, and StyleGAN, improve visual fidelity; however, such techniques usually struggle with mode collapse and unstable gradients at high network depth. This paper proposes a novel GAN structural model that incorporates deeper inception-inspired convolution and dilated convolution. This novel model is termed the Inception Generative Adversarial Network (IGAN). The IGAN model generates high-quality synthetic images while maintaining training stability, by reducing mode collapse as well as preventing vanishing and exploding gradients. Our proposed IGAN model achieves the Frechet Inception Distance (FID) of 13.12 and 15.08 on the CUB-200 and ImageNet datasets, respectively, representing a 28-33% improvement in FID over the state-of-the-art GANs. Additionally, the IGAN model attains an Inception Score (IS) of 9.27 and 68.25, reflecting improved image diversity and generation quality. Finally, the two techniques of dropout and spectral normalization are utilized in both the generator and discriminator structures to further mitigate gradient explosion and overfitting. These findings confirm that the IGAN model potentially balances training stability with image generation quality, constituting a scalable and computationally efficient framework for high-fidelity image synthesis.