Self-Attention Generative Adversarial Networks

TL;DR

SAGAN introduces self-attention mechanisms into GANs, enabling long-range dependency modeling for improved high-resolution image generation, achieving state-of-the-art results on ImageNet.

Contribution

The paper presents a novel self-attention module for GANs that captures long-range dependencies, enhancing detail generation and consistency in images.

Findings

Achieved Inception score of 52.52 on ImageNet.

Reduced FID to 18.65, outperforming previous models.

Visualization shows attention focuses on object shapes.

Abstract

In this paper, we propose the Self-Attention Generative Adversarial Network (SAGAN) which allows attention-driven, long-range dependency modeling for image generation tasks. Traditional convolutional GANs generate high-resolution details as a function of only spatially local points in lower-resolution feature maps. In SAGAN, details can be generated using cues from all feature locations. Moreover, the discriminator can check that highly detailed features in distant portions of the image are consistent with each other. Furthermore, recent work has shown that generator conditioning affects GAN performance. Leveraging this insight, we apply spectral normalization to the GAN generator and find that this improves training dynamics. The proposed SAGAN achieves the state-of-the-art results, boosting the best published Inception score from 36.8 to 52.52 and reducing Frechet Inception distance from 27.62 to 18.65 on the challenging ImageNet dataset. Visualization of the attention layers shows that the generator leverages neighborhoods that correspond to object shapes rather than local regions of fixed shape.