List your company
ResearchFormerly WizWand

Image Classification on CIFAR-100 LT Imbalance Ratio 10 (test)

91.3AccuracyLIFT
48.368859.514470.6681.8056Dec 13, 2020May 30, 2021Nov 14, 2021May 1, 2022Oct 16, 2022Apr 2, 2023Sep 18, 2023
Updated 8mo ago

Evaluation results

MethodLinks
LIFTBackbone=ViT-B/16, Learnable Params.=0.10M, Epochs=10, Training Protocol=Fine-tuning pre-trained model from ImageNet-21K2023.0991.3—
LPTBackbone=ViT-B/16, Learnable Params.=1.01M, Epochs=40+40, Training Protocol=Fine-tuning pre-trained model from ImageNet-21K2023.0991—
LIFT w/ TTEBackbone=ViT-B/16, Learnable Params.=0.10M, Epochs=10, Training Protocol=Fine-tuning pre-trained model2023.0984.9—
LIFTBackbone=ViT-B/16, Learnable Params.=0.10M, Epochs=10, Training Protocol=Fine-tuning pre-trained model2023.0983.8—
Meta-SuperDisco2023.0369.3—
LiVTBackbone=ViT-B/16, Learnable Params.=85.80M, Epochs=100, Training Protocol=Fine-tuning pre-trained model2023.0969.2—
Alshammari et al.Venue=CVPR 222023.0368.8—
Zhu et al.Venue=CVPR 222023.0364.9—
BCLBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=200, Training Protocol=Training from scratch2023.0964.9—
Cui et al.Venue=ICCV 212023.0364.2—
PaCoBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=400, Training Protocol=Training from scratch2023.0964.2—
Samuel et al.Venue=ICCV 212023.0363.4—
MiSLASBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=200+10, Training Protocol=Training from scratch2023.0963.2—
BSBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=400, Training Protocol=Training from scratch2023.0963—
Zhong et al.Venue=CVPR 212023.0362.5—
Li et al.Venue=CVPR 212023.0362.3—
DiVEBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=200, Training Protocol=Training from scratch2023.0962—
Wang et al.Venue=ICLR 212023.0361.8—
CBD + TailCalibXType=Ours2021.1161.13—
ABSGD (CB-CE)Backbone=ResNet32, Loss Function=CB-CE2020.1261.12—
CBDType=Distillation, re-implementation=true2021.1160.85—
ABSGD (CB-CE)Backbone=ResNet32, Loss Function=CB-CE2020.1260.77—
ABSGD (CB-Focal)Backbone=ResNet32, Loss Function=CB-Focal2020.1259.75—
CosineCE + TailCalibXType=Ours2021.1159.73—
GBMType=Loss or Re-weighting2021.1159.6—
META (CB-Focal)Backbone=ResNet32, Loss Function=CB-Focal2020.1259.59—
META (CB-CE)Backbone=ResNet32, Loss Function=CB-CE2020.1259.58—
Domain AdaptationType=Loss or Re-weighting2021.1159.58—
Park et al.Venue=ICCV 212023.0359.5—
SGD (CB-LDAM)Backbone=ResNet32, Loss Function=CB-LDAM2020.1259.46—
ABSGD (CB-LDAM)Backbone=ResNet32, Loss Function=CB-LDAM2020.1259.21—
ABSGD (CB-Focal)Backbone=ResNet32, Loss Function=CB-Focal2020.1259.14—
BBNType=Decouple2021.1159.12—
BBNBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=200, Training Protocol=Training from scratch2023.0959.1—
Meta Weight NetType=Loss or Re-weighting2021.1158.91—
META (CB-LDAM)Backbone=ResNet32, Loss Function=CB-LDAM2020.1258.8—
ABSGD (CB-LDAM)Backbone=ResNet32, Loss Function=CB-LDAM2020.1258.74—
SGD (CB-LDAM)Backbone=ResNet32, Loss Function=CB-LDAM2020.1258.71—
LDAM-DRWType=Loss or Re-weighting2021.1158.71—
LDAMBackbone=ResNet-32, Learnable Params.=0.46M, Epochs=200, Training Protocol=Training from scratch2023.0958.7—
CosineCE + TailCalibType=Ours2021.1158.6—
CosineCEType=Baseline, re-implementation=true2021.1158.58—
MODALSType=Generation, re-implementation=true2021.1158.53—
M2MType=Generation2021.1158.2—
mixupType=Generation2021.1158.02—
SGD (CB-Focal)Backbone=ResNet32, Loss Function=CB-Focal2020.1257.99—
Class-Balanced FocalType=Loss or Re-weighting2021.1157.99—
CRTType=Decouple, re-implementation=true2021.1157.61—
SGD (CB-CE)Backbone=ResNet32, Loss Function=CB-CE2020.1257.56—
CEType=Baseline, re-implementation=true2021.1157—
ResamplingBackbone=ResNet32, Loss Function=CE2020.1256.92—
SGD (LDAM)Backbone=ResNet32, Loss Function=LDAM2020.1256.91—
Manifold mixupType=Generation2021.1156.55—
SGD (LDAM)Backbone=ResNet32, Loss Function=LDAM2020.1256.27—
Focal LossType=Loss or Re-weighting2021.1155.78—
ResamplingBackbone=ResNet32, Loss Function=CE2020.1254.35—
SGD (CB-CE)Backbone=ResNet32, Loss Function=CB-CE2020.1253.39—
L2RWType=Loss or Re-weighting2021.1152.12—
SGD (CB-Focal)Backbone=ResNet32, Loss Function=CB-Focal2020.1250.02—
CB-CEBackbone=ResNet-32, Imbalance Ratio=102021.11—57.1
CB-FocalBackbone=ResNet-32, Imbalance Ratio=102021.11—58
CEBackbone=ResNet-32, Reproduced=true2021.04—55.71
CEBackbone=ResNet-32, Imbalance Ratio=102021.11—55.7
CE-DRSBackbone=ResNet-32, Imbalance Ratio=102021.11—56.1
CE-DRWBackbone=ResNet-32, Imbalance Ratio=102021.11—56.2
cRTBackbone=ResNet-32, Reproduced=true2021.04—57.93
DRO-LTBackbone=ResNet-32, Radius Config=ε = 0 (ERM)2021.04—59.54
DRO-LTBackbone=ResNet-32, Radius Config=Shared ε2021.04—61.22
DRO-LTBackbone=ResNet-32, Radius Config=ε/√n2021.04—63.1
DRO-LTBackbone=ResNet-32, Radius Config=Learned ε2021.04—63.41
FocalBackbone=ResNet-32, Imbalance Ratio=102021.11—55.8
Focal LossBackbone=ResNet-322021.04—55.78
KCL†Backbone=ResNet-32, Imbalance Ratio=102021.11—57.6
LDAMBackbone=ResNet-32, Imbalance Ratio=102021.11—56.9
LDAM LossBackbone=ResNet-322021.04—56.91
LDAM-DRWBackbone=ResNet-32, Imbalance Ratio=102021.11—58.7
M2m-ERMBackbone=ResNet-32, Imbalance Ratio=102021.11—58.2
M2m-LDAMBackbone=ResNet-32, Imbalance Ratio=102021.11—57.6
ReSampleBackbone=ResNet-322021.04—55.06
ReWeightBackbone=ResNet-322021.04—57.12
smDRAGONBackbone=ResNet-322021.04—58.01
SSLBackbone=ResNet-32, Reproduced=true2021.04—56.7
SSPBackbone=ResNet-322021.04—58.91
TSCBackbone=ResNet-32, Imbalance Ratio=102021.11—59
τ-normBackbone=ResNet-32, Reproduced=true2021.04—57.06