Conditional Image Generation On Imagenet

评估指标

FID

Inception score

评测结果

各个模型在此基准测试上的表现结果

			Paper Title	Repository
Projection Discriminator	27.62	36.8	cGANs with Projection Discriminator
SAGAN	18.65	52.52	Self-Attention Generative Adversarial Networks
Your Local GAN	15.94	57.22	Your Local GAN: Designing Two Dimensional Local Attention Mechanisms for Generative Models
HamGAN	14.80	58.75	Is Attention Better Than Matrix Decomposition?
FQ-GAN	13.77	54.36	Feature Quantization Improves GAN Training
BigGAN + instance selection	9.61	114.32	Instance Selection for GANs
IC-GAN + DA	9.5	108.6	Instance-Conditioned GAN
BigGAN	8.7	98.8	Large Scale GAN Training for High Fidelity Natural Image Synthesis
Omni-GAN	8.30	190.94	Omni-GAN: On the Secrets of cGANs and Beyond
ReACGAN	8.206	96.299	Rebooting ACGAN: Auxiliary Classifier GANs with Stable Training
ADC-GAN	8.02	108.10	Conditional GANs with Auxiliary Discriminative Classifier
S3 GAN	7.7	83.1	High-Fidelity Image Generation With Fewer Labels
CR-BigGAN	6.66	-	Consistency Regularization for Generative Adversarial Networks	-
Omni-INR-GAN	6.53	262.85	Omni-GAN: On the Secrets of cGANs and Beyond
BigGAN-deep	5.7	124.5	Large Scale GAN Training for High Fidelity Natural Image Synthesis
simple diffusion (U-ViT, L)	3.23	171.9	Simple diffusion: End-to-end diffusion for high resolution images
ADM-G (classifier_scale=0.5)	2.97	-	Diffusion Models Beat GANs on Image Synthesis
simple diffusion (U-Net)	2.88	137.3	Simple diffusion: End-to-end diffusion for high resolution images
ADM-G + EDS + ECT (ED-DPM, classifier_scale=0.6)	2.68	169.24	Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation
ADM-G + EDS (ED-DPM, classifier_scale=0.4)	2.63	159.72	Entropy-driven Sampling and Training Scheme for Conditional Diffusion Generation

0 of 22 row(s) selected.