HyperAI超神经

Image Generation On Celeba Hq 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

模型名称
FID
Paper TitleRepository
LFM5.26Flow Matching in Latent Space
DC-VAE15.81Dual Contradistinctive Generative Autoencoder-
UNCSN++ (RVE) + ST7.16Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation
VAEBM20.38VAEBM: A Symbiosis between Variational Autoencoders and Energy-based Models
RDUOT5.6A High-Quality Robust Diffusion Framework for Corrupted Dataset
DDGAN7.64Tackling the Generative Learning Trilemma with Denoising Diffusion GANs
WaveDiff5.94Wavelet Diffusion Models are fast and scalable Image Generators
LDM-45.11High-Resolution Image Synthesis with Latent Diffusion Models
DDMI8.73DDMI: Domain-Agnostic Latent Diffusion Models for Synthesizing High-Quality Implicit Neural Representations
VQGAN+Transformer10.2Taming Transformers for High-Resolution Image Synthesis
BOSS-Bellman Optimal Stepsize Straightening of Flow-Matching Models
RDM3.15Relay Diffusion: Unifying diffusion process across resolutions for image synthesis
Dual-MCMC EBM15.89Learning Energy-based Model via Dual-MCMC Teaching-
StyleSwin3.25StyleSwin: Transformer-based GAN for High-resolution Image Generation
LSGM7.22Score-based Generative Modeling in Latent Space
Joint-EBM9.89Learning Joint Latent Space EBM Prior Model for Multi-layer Generator-
RNODE-How to train your neural ODE: the world of Jacobian and kinetic regularization
Diffusion-JEBM8.78Learning Latent Space Hierarchical EBM Diffusion Models-
GLOW68.93Glow: Generative Flow with Invertible 1x1 Convolutions
0 of 19 row(s) selected.