List your company
ResearchFormerly WizWand

Image Classification on CIFAR-100-LT Imbalance Ratio 10

91.8Top-1 AccGNM-PT
54.3664.0873.883.52Dec 1, 2020Jul 26, 2021Mar 21, 2022Nov 14, 2022Jul 10, 2023Mar 4, 2024Oct 28, 2024
Updated 8mo ago

Evaluation results

MethodLinks
GNM-PTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.1091.8
LPTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.1091
VPTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.1089.6
LIVTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.1069.2
GML lossEpochs=400, Backbone=ResNet-322023.0567
GML lossEpochs=200, Backbone=ResNet-322023.0565.7
Difficulty-Net + Balanced SoftmaxTraining setting=e2e training, Backbone=ResNet-322022.0965.67
GPaCotrained in same setting=true2022.0965.4
GPaCoBackbone=ResNet32, Optimization Strategy (SAM)=false2024.1065.4
BS + CMOBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.1265.3
Difficulty-Net + LASTraining setting=decoupled learning, Backbone=ResNet-322022.0965.22
Difficulty-Net + LWSTraining setting=decoupled learning, Backbone=ResNet-322022.0965.08
BCLEpochs=200, Backbone=ResNet-322023.0564.9
BCLBackbone=ResNet32, Optimization Strategy (SAM)=false2024.1064.9
Difficulty-Net + cRTTraining setting=decoupled learning, Backbone=ResNet-322022.0964.8
MWN + LASTraining setting=decoupled learning, Backbone=ResNet-322022.0964.24
PaCoBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.1264.2
PaCo + Balanced Softmax (copied)Training setting=e2e training, Backbone=ResNet-32, Source=Copied from original paper2022.0964.2
PaCotrained in same setting=true2022.0964.2
PaCoEpochs=400, Backbone=ResNet-322023.0564.2
LASTraining setting=decoupled learning, Backbone=ResNet-322022.0964.18
MWN + LWSTraining setting=decoupled learning, Backbone=ResNet-322022.0964.15
PaCo + Balanced SoftmaxTraining setting=e2e training, Backbone=ResNet-322022.0964.1
MWN + cRTTraining setting=decoupled learning, Backbone=ResNet-322022.0963.99
LWSTraining setting=decoupled learning, Backbone=ResNet-322022.0963.97
cRTTraining setting=decoupled learning, Backbone=ResNet-322022.0963.74
Difficulty-NetTraining setting=e2e training, Backbone=ResNet-322022.0963.52
DRO-LT (copied)Training setting=decoupled learning, Backbone=ResNet-32, Source=Copied from original paper2022.0963.41
MISLASBackbone=ResNet-32, CMO=false2021.1263.2
MISLAS2022.0963.2
MISLASEpochs=400, Backbone=ResNet-322023.0563.2
MisLASBackbone=ResNet32, Optimization Strategy (SAM)=false2024.1063.2
PaCoTraining setting=e2e training, Backbone=ResNet-322022.0963.12
BALMSTraining setting=decoupled learning, Backbone=ResNet-322022.0963.1
Hybrid-SCEpochs=200, Backbone=ResNet-322023.0563.1
BSBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.1263
Balanced Softmaxtrained in same setting=true2022.0963
Balanced SoftmaxEpochs=400, Backbone=ResNet-322023.0563
BS + CMOBackbone=ResNet-32, CMO=true2021.1262.3
CB-DATraining setting=e2e training, Backbone=ResNet-322022.0962.16
LDAMTraining setting=e2e training, Backbone=ResNet-322022.0962.08
ResLTEpochs=200, Backbone=ResNet-322023.0562
LADE2020.1261.7
LADEBackbone=ResNet-32, CMO=false2021.1261.7
CE-DRW + CMOBackbone=ResNet-32, CMO=true2021.1261.7
Balanced Softmax2020.1261.6
Balanced Softmax (BS)Backbone=ResNet-32, CMO=false2021.1261.6
CDB-CETraining setting=e2e training, Backbone=ResNet-322022.0961.52
MWNTraining setting=e2e training, Backbone=ResNet-322022.0961.44
MetaSAug-LDAMEpochs=200, Backbone=ResNet-322023.0561.3
PC Softmax2020.1261.2
Focal LossTraining setting=e2e training, Backbone=ResNet-322022.0961.1
ResLT2022.0960.8
RIDE (3 experts) + CMOBackbone=ResNet-32, CMO=true, Number of experts=32021.1260.2
EQLTraining setting=e2e training, Backbone=ResNet-322022.0960.13
Class-BalancedTraining setting=e2e training, Backbone=ResNet-322022.0959.95
RIDE (3 experts)Backbone=ResNet-32, CMO=false, Number of experts=32021.1259.8
ImbSAMBackbone=ResNet32, Optimization Strategy (SAM)=true2024.1059.7
Causal Norm2020.1259.6
Causal NormBackbone=ResNet-32, CMO=false2021.1259.6
Causal Norm2022.0959.6
Casual modelEpochs=200, Backbone=ResNet-322023.0559.6
CE + CMOBackbone=ResNet-32, CMO=true2021.1259.5
BBNBackbone=ResNet-32, CMO=false2021.1259.2
RemixBackbone=ResNet-32, CMO=false2021.1259.2
BBN2020.1259.1
BBN2022.0959.1
BBNEpochs=200, Backbone=ResNet-322023.0559.1
BBNBackbone=ResNet32, Optimization Strategy (SAM)=false2024.1059.1
Softmax2020.1259
SSPEpochs=200, Backbone=ResNet-322023.0558.9
LDAM2020.1258.7
LDAM+DRW2022.0958.7
LDAM-DRWEpochs=200, Backbone=ResNet-322023.0558.7
LDAM-DRW + CMOBackbone=ResNet-32, CMO=true2021.1258.4
IB LossBackbone=ResNet-32, CMO=false2021.1258
CB-FocalEpochs=200, Backbone=ResNet-322023.0558
CE-DRWBackbone=ResNet-32, CMO=false2021.1257.9
LDAM-DRWBackbone=ResNet-32, CMO=false2021.1257.3
Cross Entropy (CE)Backbone=ResNet-32, CMO=false2021.1256.4
Focal Loss2020.1255.8
Focal Loss2022.0955.8
Focal lossEpochs=200, Backbone=ResNet-322023.0555.8