Image Generation On Lsun Bedroom 256 X 256

FID-10k-training-steps

评测结果

各个模型在此基准测试上的表现结果

模型名称	FID-10k-training-steps	Paper Title	Repository
SAGAN	14.0595	Generative Adversarial Transformers	-
StyleGAN	-	-	-
SS-GAN (sBN)	-	Self-Supervised GANs via Auxiliary Rotation Loss	-
LFM	-	Flow Matching in Latent Space	-
Patch Diffusion	-	-	-
ADM (dropout)	-	Diffusion Models Beat GANs on Image Synthesis	-
StyleGAN (DINOv2)	-	-	-
PGGAN	-	Progressive Growing of GANs for Improved Quality, Stability, and Variation	-
Unleashing Transformers	-	Unleashing Transformers: Parallel Token Prediction with Discrete Absorbing Diffusion for Fast High-Resolution Image Generation from Vector-Quantized Codes	-
Denoising Diffusion Probabilistic Model (large)	-	Denoising Diffusion Probabilistic Models	-
StyleGAN2	11.5255	Generative Adversarial Transformers	-
Denoising Diffusion Probabilistic Model (large, DINOv2)	-	Denoising Diffusion Probabilistic Models	-
iDDPM (DINOv2)	-	-	-
TDPM+ (TTrunc=99)	-	Truncated Diffusion Probabilistic Models and Diffusion-based Adversarial Auto-Encoders	-
VQGAN	59.6333	Generative Adversarial Transformers	-
Consistency	-	-	-
Projected GAN (DINOv2)	-	Projected GANs Converge Faster	-
StackGAN-v2	-	StackGAN++: Realistic Image Synthesis with Stacked Generative Adversarial Networks	-
Projected GAN	1.52	Projected GANs Converge Faster	-
Consistency (DINOv2)	-	-	-

0 of 32 row(s) selected.