List your company
ResearchFormerly WizWand

Class-Incremental Learning on ImageNet-100 20 tasks

66.36Average AccuracySSIL + TBBN
54.50457.58260.6663.738Jan 29, 2022
Updated 4mo ago

Evaluation results

MethodLinks
SSIL + TBBNBase CIL Method=SSIL, Normalization/Block Variant=TBBN, Backbone=ResNet-342022.0166.36——
SSIL + BNBase CIL Method=SSIL, Normalization/Block Variant=BN, Backbone=ResNet-342022.0165.27——
FT + TBBNBase CIL Method=FT, Normalization/Block Variant=TBBN, Backbone=ResNet-342022.0164.76——
SSIL + TBBNBase CIL Method=SSIL, Normalization/Block Variant=TBBN, Backbone=MobileNet-v22022.0163.37——
SSIL + BNBase CIL Method=SSIL, Normalization/Block Variant=BN, Backbone=MobileNet-v22022.0162.25——
FT + BNBase CIL Method=FT, Normalization/Block Variant=BN, Backbone=ResNet-342022.0162——
FT + TBBNBase CIL Method=FT, Normalization/Block Variant=TBBN, Backbone=MobileNet-v22022.0160.41——
SSIL + TBBNBase CIL Method=SSIL, Normalization/Block Variant=TBBN, Backbone=MnasNet (x0.5)2022.0160.24——
SSIL + TBBNBase CIL Method=SSIL, Normalization/Block Variant=TBBN, Backbone=ShuffleNet-V22022.0160.07——
SSIL + BNBase CIL Method=SSIL, Normalization/Block Variant=BN, Backbone=MnasNet (x0.5)2022.0159.81——
SSIL + BNBase CIL Method=SSIL, Normalization/Block Variant=BN, Backbone=ShuffleNet-V22022.0159.69——
FT + TBBNBase CIL Method=FT, Normalization/Block Variant=TBBN, Backbone=ShuffleNet-V22022.0158.17——
FT + BNBase CIL Method=FT, Normalization/Block Variant=BN, Backbone=MobileNet-v22022.0157.43——
FT + TBBNBase CIL Method=FT, Normalization/Block Variant=TBBN, Backbone=MnasNet (x0.5)2022.0157.09——
FT + BNBase CIL Method=FT, Normalization/Block Variant=BN, Backbone=ShuffleNet-V22022.0155.37——
FT + BNBase CIL Method=FT, Normalization/Block Variant=BN, Backbone=MnasNet (x0.5)2022.0154.96——
CLA-E2026.06—45.5251.03
CLA-R2026.06—42.8649.89
CLIMB2026.06—47.4652.92
i.i.d.2026.06——53.93
MinRed2026.06—35.8743.34
Osiris-R2026.06—37.0642.72
PCMCpretraining=false2026.06—40.1139.42
SCALE2026.06—28.733.43