List your company
ResearchFormerly WizWand

Image Denoising on SIDD 1 (val)

39.43PSNRDANet
23.029227.287131.54535.8029Mar 22, 2022Jul 12, 2022Nov 2, 2022Feb 22, 2023Jun 15, 2023Oct 5, 2023Jan 26, 2024
Updated 8mo ago

Evaluation results

MethodLinks
DANetSupervision Paradigm=Supervised (Real pairs)2022.0339.430.956
VDNSupervision Paradigm=Supervised (Real pairs)2022.0339.260.955
AINDNet (R)Supervision Paradigm=Supervised (Real pairs), Self-ensemble Strategy=true2022.0338.840.951
AP-BSN + R3Supervision Paradigm=Self-supervised, Training Protocol=trained on SIDD benchmark images in a fully self-supervised fashion2022.0336.910.931
AP-BSN + R3Supervision Paradigm=Self-supervised2022.0335.970.925
C2N + DIDNSupervision Paradigm=Unsupervised (Unpaired), Self-ensemble Strategy=true2022.0335.350.937
DnCNNSupervision Paradigm=Supervised (Real pairs), Evaluation Source=evaluated by authors2022.0335.130.896
AP-BSNSupervision Paradigm=Self-supervised2022.0334.90.9
R2RSupervision Paradigm=Self-supervised2022.0334.780.898
OursAvg. Infer. time (s)=37.2, Iterations=800, EMA weight=0.992024.0134.430.844
Zhou et al.Supervision Paradigm=Supervised (Synthetic pairs), Evaluation Source=evaluated by authors2022.03340.898
DIPAvg. Infer. time (s)=333.22024.0133.680.802
Ours (faster)Avg. Infer. time (s)=7.9, Iterations=200, EMA weight=0.92024.0133.680.828
FasterDIPAvg. Infer. time (s)=138.22024.0133.550.795
CBDNetSupervision Paradigm=Supervised (Synthetic pairs)2022.0333.280.868
Noise2SelfSupervision Paradigm=Self-supervised, Evaluation Source=reported from R2R [32]2022.0329.560.808
Noise2VoidSupervision Paradigm=Self-supervised, Evaluation Source=reported from R2R [32]2022.0327.680.668
N2S*Avg. Infer. time (s)=114.42024.0126.780.573
N2V*Avg. Infer. time (s)=98.12024.0126.740.627
WNNMSupervision Paradigm=Non-learning based2022.0325.780.809
BM3DSupervision Paradigm=Non-learning based2022.0325.650.685
ZS-N2NAvg. Infer. time (s)=15.12024.0125.590.422
DnCNNSupervision Paradigm=Supervised (Synthetic pairs)2022.0323.660.583