Image Generation On Celeba 64X64

FID

評価結果

このベンチマークにおける各モデルのパフォーマンス結果

		Paper Title	Repository
Feature Alignment	128.35	Feature Alignment as a Generative Process
PAE	49.2	Probabilistic Autoencoder
FDP(MLP)	35	Continuous-Time Functional Diffusion Processes
BEGAN-CS	34.136	Escaping from Collapsing Modes in a Constrained Space
PR-BigGAN - Precision	22.45	-	-
PeerGAN	13.95	DuelGAN: A Duel Between Two Discriminators Stabilizes the GAN Training	-
CLR-GAN	13.63	CLR-GAN: Improving GANs Stability and Quality via Consistent Latent Representation and Reconstruction	-
TransGAN-XL	12.23	TransGAN: Two Pure Transformers Can Make One Strong GAN, and That Can Scale Up
FCE	12.21	Flow Contrastive Estimation of Energy-Based Models
FDP(UViT)	11	Continuous-Time Functional Diffusion Processes
HDCGAN	8.44	High-resolution Deep Convolutional Generative Adversarial Networks
PR-BigGAN - Recall	6.03	-	-
EBM-DRL (T6)	5.98	Learning Energy-Based Models by Diffusion Recovery Likelihood
C^3G-NeRF	5.6	Class-Continuous Conditional Generative Neural Radiance Field
NCP-VAE	5.25	A Contrastive Learning Approach for Training Variational Autoencoder Priors	-
DiffuseVAE	3.97	DiffuseVAE: Efficient, Controllable and High-Fidelity Generation from Low-Dimensional Latents
BigGAN-OBRS	3.74	Optimal Budgeted Rejection Sampling for Generative Models	-
Blackout Diffusion	3.22	Blackout Diffusion: Generative Diffusion Models in Discrete-State Spaces
DDPM++ (VP, NLL) + ST	2.9	Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation
PNDM	2.71	Pseudo Numerical Methods for Diffusion Models on Manifolds

0 of 39 row(s) selected.