List your company
ResearchFormerly WizWand

Noise Recovery on Linear-Gaussian distractor environment (train)

0.656Noise R^2VAE
-0.022080.153960.330.50604Jan 20, 2026
Updated 8mo ago

Evaluation results

MethodLinks
VAEScale (SNR)=8.0 (-2.2 dB), Training Time=12.3s2026.010.656
VAEScale (SNR)=4.0 (3.8 dB), Training Time=12.5s2026.010.512
ARScale (SNR)=8.0 (-2.2 dB), Training Time=7.1s2026.010.491
ARScale (SNR)=4.0 (3.8 dB), Training Time=6.5s2026.010.394
VJEPAScale (SNR)=8.0 (-2.2 dB), Training Time=13.4s2026.010.299
BJEPAScale (SNR)=8.0 (-2.2 dB), Training Time=23.0s2026.010.292
JEPAScale (SNR)=8.0 (-2.2 dB), Training Time=16.1s2026.010.226
BJEPAScale (SNR)=4.0 (3.8 dB), Training Time=23.2s2026.010.213
VJEPAScale (SNR)=4.0 (3.8 dB), Training Time=13.8s2026.010.025
JEPAScale (SNR)=4.0 (3.8 dB), Training Time=16.6s2026.010.004