Projected GAN

Projected GANs improve the training stability and efficiency of Generative Adversarial Networks (GANs) by leveraging pre-trained feature networks to guide the discriminator's learning. Current research focuses on optimizing Projected GAN architectures, such as integrating StyleGAN features or employing depth-separable convolutions, to enhance image generation speed, reduce computational costs, and improve the quality of generated images across diverse datasets, including those with limited samples. This approach holds significant promise for accelerating GAN training and expanding the applicability of GANs to various domains, from image synthesis to audio generation, by mitigating the challenges associated with traditional GAN training.

Papers

January 23, 2024

Faster Projected GAN: Towards Faster Few-Shot Image Generation
Chuang Wang, Zhengping Li, Yuwen Hao, Lijun Wang, Xiaoxue Li
Image Generation GAN Model Projected GAN

July 30, 2023

Stylized Projected GAN: A Novel Architecture for Fast and Realistic Image Generation
Md Nurul Muttakin, Malik Shahid Sultan, Robert Hoehndorf, Hernando Ombao
Generative Adversarial Network Image Generation GAN Model Novel Architecture Projected GAN

May 17, 2023

Learning Pose Image Manifolds Using Geometry-Preserving GANs and Elasticae
Shenyuan Liang, Pavan Turaga, Anuj Srivastava
GAN Model Manifold Learning Image Manifold Elastic Band Motion Manifold Projected GAN

September 5, 2022

Exploiting Pre-trained Feature Networks for Generative Adversarial Networks in Audio-domain Loop Generation
Yen-Tung Yeh, Bo-Yu Chen, Yi-Hsuan Yang
Generative Adversarial Network GAN Model GAN Training Loop Generation Projected GAN

February 1, 2022

StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets
Axel Sauer, Katja Schwarz, Andreas Geiger
Supervised ImageNet Generative Modeling StyleGAN Latent High Resolution Image Synthesis State of the Art Datasets Recent StyleGAN3 Architecture Projected GAN

Projected GAN

Papers

Faster Projected GAN: Towards Faster Few-Shot Image Generation

Stylized Projected GAN: A Novel Architecture for Fast and Realistic Image Generation

Learning Pose Image Manifolds Using Geometry-Preserving GANs and Elasticae

Exploiting Pre-trained Feature Networks for Generative Adversarial Networks in Audio-domain Loop Generation

StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets