Image Generation On Imagenet 512X512

Metrics

FID

Inception score

Results

Performance results of various models on this benchmark

Model Name	FID	Inception score	Paper Title	Repository
DiT-XL/2	3.04	240.82	Scalable Diffusion Models with Transformers
MAR-L, Diff Loss	1.73	-	Autoregressive Image Generation without Vector Quantization
SIMS	1.73	-	Self-Improving Diffusion Models with Synthetic Data	-
EDM2- S Autoguidance (XS, T /16)	1.34	-	Guiding a Diffusion Model with a Bad Version of Itself
SiD-EDM2-M (498M)	2.06	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
MAGVIT-v2 (w/o guidance)	3.07	213.1	Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation
Poly-INR	3.81	-	Polynomial Implicit Neural Representations For Large Diverse Datasets
SiDA-EDM2-M (498M)	1.488	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
PaGoDA	1.80	-	PaGoDA: Progressive Growing of a One-Step Generator from a Low-Resolution Diffusion Teacher
SiDA-EDM2-L (777M)	1.413	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
SiD-EDM2-XS (125M)	3.353	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
MaskGIT (a=0.05)	4.46	342.0	MaskGIT: Masked Generative Image Transformer
MAGVIT-v2	1.91	324.3	Language Model Beats Diffusion -- Tokenizer is Key to Visual Generation
ADM-G	7.72	172.71	Diffusion Models Beat GANs on Image Synthesis
SiDA-EDM2-XL (1.1B)	1.379	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
SiD-EDM2-S (280M)	2.707	-	Adversarial Score identity Distillation: Rapidly Surpassing the Teacher in One Step
TiTok-L-64	2.49	-	An Image is Worth 32 Tokens for Reconstruction and Generation
DPC-U	3.54	350.2	Discrete Predictor-Corrector Diffusion Models for Image Synthesis	-
GMem	1.71	-	Generative Modeling with Explicit Memory
Latent Diffusion (LDM-4-G)	3.60	247.67	High-Resolution Image Synthesis with Latent Diffusion Models

0 of 48 row(s) selected.