Image Generation On Imagenet 256X256

评估指标

FID

评测结果

各个模型在此基准测试上的表现结果

		Paper Title	Repository
Improved DDPM	12.3	Improved Denoising Diffusion Probabilistic Models
ADM	11.84	-	-
BigGAN-deep	8.1	Large Scale GAN Training for High Fidelity Natural Image Synthesis
Polarity-BigGAN	6.82	Polarity Sampling: Quality and Diversity Control of Pre-Trained Generative Networks via Singular Values
VQGAN+Transformer (k=mixed, p=1.0, a=0.005)	6.59	Taming Transformers for High-Resolution Image Synthesis
MaskGIT	6.18	MaskGIT: Masked Generative Image Transformer
VQGAN+Transformer (k=600, p=1.0, a=0.05)	5.2	Taming Transformers for High-Resolution Image Synthesis
CDM	4.88	Cascaded Diffusion Models for High Fidelity Image Generation	-
ADM-G	4.59	Diffusion Models Beat GANs on Image Synthesis
RIN	4.51	Scalable Adaptive Computation for Iterative Generation
LFM	4.46	Flow Matching in Latent Space
ADM-G++ (Recall)	4.45	Refining Generative Process with Discriminator Guidance in Score-based Diffusion Models
LDM	4.29	-	-
ADM-G + EDS + ECT (ED-DPM, classifier_scale=1.0)	4.09	Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation
MaskGIT (a=0.05)	4.02	MaskGIT: Masked Generative Image Transformer
ADM-G + EDS (ED-DPM, classifier_scale=0.75)	3.96	Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation
ADM-G, ADM-U	3.94	Diffusion Models Beat GANs on Image Synthesis
RQ-Transformer	3.83	Autoregressive Image Generation using Residual Quantization
simple diffusion (U-ViT, L)	3.75	Simple diffusion: End-to-end diffusion for high resolution images
simple diffusion (U-Net)	3.71	Simple diffusion: End-to-end diffusion for high resolution images

0 of 79 row(s) selected.