List your company
ResearchFormerly WizWand

Unconditional Image Generation on CIFAR-10 (test)

1.54FIDGDD-I
1.399842.345923.2924.23808Jun 19, 2020May 25, 2021Apr 30, 2022Apr 6, 2023Mar 11, 2024Feb 14, 2025Jan 21, 2026
Updated 6mo ago

Evaluation results

MethodLinks
GDD-INFE=12024.051.5410.1——
GDDNFE=12024.051.6610.11——
EDM-G++Space=Data, NFE=352022.111.77—2.55—
CTMNFE=22024.051.87———
PFGM++NFE=35, # param. (M)=55.72024.101.91———
SiDNFE=1, alpha=1.22024.051.929.98——
SiDNFE=1, # param. (M)=55.72024.101.92———
LSGM-G++Space=Latent, NFE=1382022.111.94—3.42—
EDMSpace=Data, NFE=35, Seed=manual2022.111.97—2.6—
EDMNFE=35, # param. (M)=55.72024.101.97———
EDMNFE=352024.051.98———
CTMNFE=12024.051.98———
SiDNFE=1, alpha=1.02024.052.0210.01——
EDMSpace=Data, NFE=39, Seed=random2022.112.03—2.6—
TCMNFE=2, # param. (M)=55.7, Steps=2-step2024.102.05———
LSGMSpace=Latent, NFE=1382022.112.1—3.43—
LSGMNFE=147, # param. (M)=4752024.102.1———
ECMNFE=2, # param. (M)=55.7, Steps=2-step2024.102.11———
NCSN++Continuity Objective=continuous, SDE Type=VE, Architecture Depth=deep2020.112.29.89——
NCSN++ cont.2021.072.29.89——
NCSN++Space=Data, NFE=20002022.112.2—3.45—
iCT-deepNFE=22024.052.249.89——
iCT-deepNFE=2, # param. (M)=112, Steps=2-step2024.102.24———
CLD-SGMSpace=Data, NFE=3122022.112.25—3.31—
INDMSpace=Latent, NFE=20002022.112.28—3.09—
NCSN++Continuity Objective=continuous, SDE Type=VE, Architecture Depth=standard2020.112.389.83——
DDPM++Continuity Objective=continuous, SDE Type=VP, Architecture Depth=deep2020.112.419.68——
DDPM++Continuity Objective=continuous, SDE Type=sub-VP, Architecture Depth=deep2020.112.419.57——
NCSN++Continuity Objective=discrete, Architecture Depth=standard2020.112.459.73——
iCTNFE=22024.052.469.8——
TCMNFE=1, # param. (M)=55.7, Steps=1-step2024.102.46———
iCTNFE=2, # param. (M)=56.4, Steps=2-step2024.102.46———
Soft TruncationSpace=Data, NFE=20002022.112.47—2.91—
iCT-deepNFE=12024.052.519.76——
iCT-deepNFE=1, # param. (M)=112, Steps=1-step2024.102.51———
DDPM++Continuity Objective=continuous, SDE Type=VP, Architecture Depth=standard2020.112.559.58——
DDPM++Continuity Objective=continuous, SDE Type=sub-VP, Architecture Depth=standard2020.112.619.56——
DMDNFE=12024.052.62———
DDPM++Continuity Objective=discrete, Architecture Depth=standard2020.112.789.64——
iCTNFE=12024.052.839.54——
iCTNFE=1, # param. (M)=56.4, Steps=1-step2024.102.83———
Improved DDPMloss=Lsimple2021.072.9———
iDDPMSpace=Data, NFE=10002022.112.9—3.37—
StyleGAN2-ADAMode=Unconditional2020.112.929.83——
DDPM++ cont. (deep, sub-VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=sub-VP, Network Depth=Deep2020.112.92—2.99—
DDPM++ cont2021.072.92—2.99—
StyleGAN2-ADANFE=12024.052.929.82——
CDNFE=22024.052.93———
DDPM++ cont. (deep, VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=VP, Network Depth=Deep2020.113.08—3.13—
DDPM++ cont. (sub-VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=sub-VP2020.113.16—3.02—
Denoising Diffusion Probabilistic ModelsObjective=L_simple2020.063.179.463.75—
DDPMMode=Unconditional2020.113.179.46——
DDPM (Lsimple)Likelihood Type=ELBO, Training Objective=Discrete2020.113.17—3.75—
DDPMloss=Lsimple2021.073.179.46——
DDPMSpace=Data, NFE=10002022.113.17—3.75—
DDPMNFE=10002024.053.17———
DDPMNFE=1000, # param. (M)=35.72024.103.17———
StyleGAN2 + ADA (v1)2020.063.269.74——
StyleGAN2 + ADAAugmentation=Adaptive Discriminator Augmentation2021.073.269.74——
TRACTNFE=2, # param. (M)=55.72024.103.32———
DDPMLikelihood Type=Exact ODE Likelihood, Training Objective=Discrete2020.113.37—3.28—
DeMe (After Merge)#Iterations=20Kx42024.103.51———
CDNFE=12024.053.55———
DDPM cont. (sub-VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=sub-VP2020.113.56—3.05—
ECMNFE=1, # param. (M)=55.7, Steps=1-step2024.103.6———
GENIENFEs=25, Score model checkpoint=Same, Striding schedule search=true2022.103.64———
GENIENFEs=25, Score model checkpoint=Same, Striding schedule search=false2022.103.67———
DDPM cont. (VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=VP2020.113.69—3.21—
DMDNFE=1, # param. (M)=55.72024.103.77———
DFNONFE=12024.053.78———
DSNONFE=1, # param. (M)=65.82024.103.78———
TRACTNFE=1, # param. (M)=55.72024.103.78———
DeMe (Before Merge)#Iterations=20Kx42024.103.79———
DDPM++ cont. (VP)Likelihood Type=Exact ODE Likelihood, Training Objective=Continuous, SDE Formulation=VP2020.113.93—3.16—
GENIENFEs=20, Score model checkpoint=Same, Striding schedule search=true2022.103.94———
GENIENFEs=20, Score model checkpoint=Same, Striding schedule search=false2022.103.94———
DDIMDiffusion steps=1002021.064.16———
DDIMNFE=1002024.054.16———
ANT-UW#Iterations=80K2024.104.21———
ANT-NashMTL#Iterations=80K2024.104.24———
Learned SamplerNFEs=25, Score model checkpoint=Different, Striding schedule search=true2022.104.25———
Uniform-Reflowtype=reflow2023.124.33———
Distill-6-Reflowsteps=62023.124.35———
Before-finetuning2024.104.42———
Learned SamplerNFEs=25, Score model checkpoint=Different, Striding schedule search=false2022.104.47———
GENIENFEs=15, Score model checkpoint=Same, Striding schedule search=true2022.104.49———
GENIENFEs=15, Score model checkpoint=Same, Striding schedule search=false2022.104.49———
Trun-SNR#Iterations=80K2024.104.49———
S-PNDMNFEs=25, Score model checkpoint=Same2022.104.51———
PDNFE=2, # param. (M)=602024.104.51———
Diff-InstructNFE=1, # param. (M)=55.72024.104.53———
StyleGAN2-D + ViTGAN-GDiscriminator=StyleGAN2, Generator=ViTGAN2021.074.579.89——
Learned SamplerNFEs=20, Score model checkpoint=Different, Striding schedule search=true2022.104.72———
F-PNDMNFEs=25, Score model checkpoint=Same2022.104.73———
BOSS-6steps=62023.124.8———
Ambient DataLoops (Loop 1)Corruption Type=JPEG, q=50, rho=1.2/2^32026.014.825———
Learned SamplerNFEs=20, Score model checkpoint=Different, Striding schedule search=false2022.104.89———
ViTGAN2021.074.929.69——
Ambient DataLoops (Loop 1)Corruption Type=Blur, sigma_B=0.6, rho=1.2/2^32026.014.947———
Ambient DataLoops (Loop 1)Corruption Type=Blur, sigma_B=0.8, rho=1.9/2^32026.015.044———