Image Generation On Ffhq 1024 X 1024

FID

Results

Performance results of various models on this benchmark

Model Name	FID	Paper Title
StyleGAN3-T	2.79	Alias-Free Generative Adversarial Networks
Diffusion StyleGAN2	2.83	Diffusion-GAN: Training GANs with Diffusion
Very Deep VAE	-	Very Deep VAEs Generalize Autoregressive Models and Can Outperform Them on Images
Efficient-VDVAE	-	Efficient-VDVAE: Less is more
StyleALAE	13.09	Adversarial Latent Autoencoders
SWAGAN-Bi	4.06	SWAGAN: A Style-based Wavelet-driven Generative Model
MSG-StyleGAN	5.8	MSG-GAN: Multi-Scale Gradients for Generative Adversarial Networks
Polarity-StyleGAN2	2.57	Polarity Sampling: Quality and Diversity Control of Pre-Trained Generative Networks via Singular Values
StyleGAN	4.4	A Style-Based Generator Architecture for Generative Adversarial Networks
StyleNAT	4.17	StyleNAT: Giving Each Head a New Perspective
StyleGAN3-R	3.07	Alias-Free Generative Adversarial Networks
StyleGAN2	2.84	Analyzing and Improving the Image Quality of StyleGAN
StyleSwin	5.07	StyleSwin: Transformer-based GAN for High-resolution Image Generation
StyleSAN-XL	1.61	SAN: Inducing Metrizability of GAN with Discriminative Normalized Linear Layer
CIPS	10.07	Image Generators with Conditionally-Independent Pixel Synthesis
MaGNET-StyleGAN2	2.66	MaGNET: Uniform Sampling from Deep Generative Network Manifolds Without Retraining
StyleGAN-XL	2.02	StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets
FQ-GAN	3.19	Feature Quantization Improves GAN Training
HiT-B	6.37	Improved Transformer for High-Resolution GANs
StyleGAN2 ADA+bCR	3.62	Training Generative Adversarial Networks with Limited Data

0 of 20 row(s) selected.