Image Generation On Imagenet 256X256

Metrics

FID

Results

Performance results of various models on this benchmark

Model Name	FID	Paper Title	Repository
simple diffusion (U-Net)	3.71	Simple diffusion: End-to-end diffusion for high resolution images	-
RDM	1.99	Relay Diffusion: Unifying diffusion process across resolutions for image synthesis	-
Patch Diffusion	2.74	-	-
RAR-L, autoregressive	1.70	Randomized Autoregressive Visual Generation	-
RAR-B, autoregressive	1.95	Randomized Autoregressive Visual Generation	-
RAR-XL, autoregressive	1.50	Randomized Autoregressive Visual Generation	-
CDM	4.88	Cascaded Diffusion Models for High Fidelity Image Generation	-
MDT	1.79	MDTv2: Masked Diffusion Transformer is a Strong Image Synthesizer	-
TiTok-S-128	1.97	An Image is Worth 32 Tokens for Reconstruction and Generation	-
BIGRoC-pl (Guided-Diffusion)	3.69	BIGRoC: Boosting Image Generation via a Robust Classifier	-
ADM-G++ (Recall)	4.45	Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models	-
GigaGAN	3.45	Scaling up GANs for Text-to-Image Synthesis	-
MaskGIT	6.18	MaskGIT: Masked Generative Image Transformer	-
SiT-XL/2 + REPA (with the guidance interval)	1.42	Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think	-
Discriminator Guidance	1.83	Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models	-
SiT-XL/2 + MG	1.34	Diffusion Models without Classifier-free Guidance	-
xAR-L	1.28	Beyond Next-Token: Next-X Prediction for Autoregressive Visual Generation	-
ADM-G + EDS (ED-DPM, classifier_scale=0.75)	3.96	Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation	-
StyleGAN-XL	2.30	StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets	-
RQ-Transformer	3.83	Autoregressive Image Generation using Residual Quantization	-

0 of 79 row(s) selected.