List your company
ResearchFormerly WizWand

Image Classification on CIFAR-100-LT Imbalance Ratio 100

0.903Top-1 AccGNM-PT
0.363240.503370.64350.78363Dec 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)=false2024.100.903
LPTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.100.891
VPTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.100.81
LIVTBackbone=ViT-B/16, Optimization Strategy (SAM)=false2024.100.582
SHIKEBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.563
ImbSAMBackbone=ResNet32, Optimization Strategy (SAM)=true2024.100.548
NCLBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.542
GML lossEpochs=400, Backbone=ResNet-322023.050.54
GML lossEpochs=200, Backbone=ResNet-322023.050.53
Difficulty-Net + LASTraining setting=decoupled learning, Backbone=ResNet-322022.090.5296
Difficulty-Net + LWSTraining setting=decoupled learning, Backbone=ResNet-322022.090.5262
GPaCotrained in same setting=true2022.090.523
GPaCoBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.523
Difficulty-Net + Balanced SoftmaxTraining setting=e2e training, Backbone=ResNet-322022.090.5214
Difficulty-Net + cRTTraining setting=decoupled learning, Backbone=ResNet-322022.090.5201
PaCoBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.120.52
PaCo + Balanced Softmax (copied)Training setting=e2e training, Backbone=ResNet-32, Source=Copied from original paper2022.090.52
PaCotrained in same setting=true2022.090.52
PaCoEpochs=400, Backbone=ResNet-322023.050.52
BCLEpochs=200, Backbone=ResNet-322023.050.519
BCLBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.519
BS + CMOBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.120.517
PaCo + Balanced SoftmaxTraining setting=e2e training, Backbone=ResNet-322022.090.5147
BALMSTraining setting=decoupled learning, Backbone=ResNet-322022.090.5095
BSBackbone=ResNet-32, Training Epochs=400, Augmentation=AutoAugment [9]2021.120.508
Balanced Softmaxtrained in same setting=true2022.090.508
Balanced SoftmaxEpochs=400, Backbone=ResNet-322023.050.508
CCSAMBackbone=ResNet32, Optimization Strategy (SAM)=true2024.100.508
RIDEBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.504
RIDE (3 experts) + CMOBackbone=ResNet-32, CMO=true, Number of experts=32021.120.5
MWN + LASTraining setting=decoupled learning, Backbone=ResNet-322022.090.4912
LASTraining setting=decoupled learning, Backbone=ResNet-322022.090.4868
MWN + LWSTraining setting=decoupled learning, Backbone=ResNet-322022.090.4865
RIDE (3 experts)Backbone=ResNet-32, CMO=false, Number of experts=32021.120.486
GCLBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.486
MWN + cRTTraining setting=decoupled learning, Backbone=ResNet-322022.090.4834
ResLTEpochs=200, Backbone=ResNet-322023.050.482
LWSTraining setting=decoupled learning, Backbone=ResNet-322022.090.4813
cRTTraining setting=decoupled learning, Backbone=ResNet-322022.090.4804
MetaSAug-LDAMEpochs=200, Backbone=ResNet-322023.050.48
Difficulty-NetTraining setting=e2e training, Backbone=ResNet-322022.090.4796
DRO-LT (copied)Training setting=decoupled learning, Backbone=ResNet-32, Source=Copied from original paper2022.090.4731
PaCoTraining setting=e2e training, Backbone=ResNet-322022.090.4726
LDAM-DRW + CMOBackbone=ResNet-32, CMO=true2021.120.472
MISLASBackbone=ResNet-32, CMO=false2021.120.47
CE-DRW + CMOBackbone=ResNet-32, CMO=true2021.120.47
MISLAS2022.090.47
MISLASEpochs=400, Backbone=ResNet-322023.050.47
MisLASBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.47
Hybrid-SCEpochs=200, Backbone=ResNet-322023.050.467
BS + CMOBackbone=ResNet-32, CMO=true2021.120.466
EQLTraining setting=e2e training, Backbone=ResNet-322022.090.4647
CB-DATraining setting=e2e training, Backbone=ResNet-322022.090.4624
LDAMTraining setting=e2e training, Backbone=ResNet-322022.090.4614
RemixBackbone=ResNet-32, CMO=false2021.120.458
LADE2020.120.454
LADEBackbone=ResNet-32, CMO=false2021.120.454
PC Softmax2020.120.453
ResLT2022.090.453
CDB-CETraining setting=e2e training, Backbone=ResNet-322022.090.4525
Balanced Softmax2020.120.451
Balanced Softmax (BS)Backbone=ResNet-32, CMO=false2021.120.451
IB LossBackbone=ResNet-32, CMO=false2021.120.45
MWNTraining setting=e2e training, Backbone=ResNet-322022.090.4481
Class-BalancedTraining setting=e2e training, Backbone=ResNet-322022.090.4478
Causal Norm2020.120.441
Causal NormBackbone=ResNet-32, CMO=false2021.120.441
Causal Norm2022.090.441
Casual modelEpochs=200, Backbone=ResNet-322023.050.441
Focal LossTraining setting=e2e training, Backbone=ResNet-322022.090.4403
CE + CMOBackbone=ResNet-32, CMO=true2021.120.439
SSPEpochs=200, Backbone=ResNet-322023.050.434
BBN2020.120.426
BBNBackbone=ResNet-32, CMO=false2021.120.426
BBN2022.090.426
BBNEpochs=200, Backbone=ResNet-322023.050.426
BBNBackbone=ResNet32, Optimization Strategy (SAM)=false2024.100.426
LDAM2020.120.42
LDAM+DRW2022.090.42
LDAM-DRWEpochs=200, Backbone=ResNet-322023.050.42
LDAM-DRWBackbone=ResNet-32, CMO=false2021.120.417
CE-DRWBackbone=ResNet-32, CMO=false2021.120.411
Softmax2020.120.41
CB-FocalEpochs=200, Backbone=ResNet-322023.050.396
Cross Entropy (CE)Backbone=ResNet-32, CMO=false2021.120.386
Focal Loss2020.120.384
Focal Loss2022.090.384
Focal lossEpochs=200, Backbone=ResNet-322023.050.384