List your company
ResearchFormerly WizWand

Image Classification on STL-10 standard (test)

19.25Test Error (%)VGG8B+CO (predsim)
18.696822.430926.16529.8991Jan 20, 2019
Updated 8mo ago

Evaluation results

MethodLinks
VGG8B+CO (predsim)Backbone=VGG8B, Parameters=12M, Loss Function=Predsim Loss (pred+sim), Regularization=Cutout2019.0119.25
VGG8B (predsim)Backbone=VGG8B, Parameters=12M, Loss Function=Predsim Loss (pred+sim)2019.0120.51
WideResNet-16-8+CO (global)Backbone=WideResNet-16-8, Parameters=11M, Loss Function=Global Back-propagation, Regularization=Cutout2019.0120.77
VGG8B (sim)Backbone=VGG8B, Parameters=12M, Loss Function=Local Similarity Matching Loss (sim)2019.0123.15
WideResNet-16-8 (global)Backbone=WideResNet-16-8, Parameters=11M, Loss Function=Global Back-propagation2019.0123.48
VGG8B (pred)Backbone=VGG8B, Parameters=12M, Loss Function=Local Prediction Loss (pred)2019.0126.83
VGG8B (global)Backbone=VGG8B, Parameters=12M, Loss Function=Global Back-propagation2019.0133.08