Image Classification on Clothing1M (test)
82.1AccuracySOMNet
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SOMNetBackbone=ResNet-502019.06 | 82.1 | — | — | |
| 50-ResNet (Fine-tuned)loss=cross-entropy, init=#6, training=50k2016.09 | 80.38 | — | — | |
| Nonedata=50k clean, pretrained=#3 model2017.11 | 80.38 | — | — | |
| Robust Self-Training with Closed-loop Label CorrectionExtra data=Meta-knowledge2026.03 | 80.23 | — | — | |
| Robust Self-Training with Closed-loop Label CorrectionExtra data=Meta-knowledge, Clean data percentage=50%2026.03 | 79.92 | — | — | |
| Nonedata=50k clean, pretrained=#6 model2017.11 | 79.9 | — | — | |
| CurriculumNetBackbone=ResNet-50, Source=Re-run2019.06 | 79.8 | — | — | |
| EMLCExtra Data=true, k=12023.05 | 79.35 | — | — | |
| EMLCExtra data=Meta-knowledge, Self-supervised learning=true2026.03 | 79.35 | — | — | |
| AlexNetloss=cross-entropy, init=#1, training=1M, 50k2016.09 | 78.24 | — | — | |
| Robust Self-Training with Closed-loop Label CorrectionExtra data=Meta-knowledge, Clean data percentage=10%2026.03 | 78.17 | — | — | |
| EMLCExtra Data=true, k=52023.05 | 78.16 | — | — | |
| FasTENClean set usage=true, Loss Correction=true, Backbone=ResNet-50, Pre-trained=ImageNet2021.11 | 77.83 | — | — | |
| FasTENExtra Data=true2023.05 | 77.83 | — | — | |
| FasTENExtra data=Direct training on clean instances2026.03 | 77.83 | — | — | |
| PUDistillBackbone=ResNet-50, Pre-trained=ImageNet2021.05 | 77.7 | — | — | |
| L2B-Naive2022.02 | 77.5 | — | — | |
| DMLPExtra data=Meta-knowledge, Self-supervised learning=true2026.03 | 77.31 | — | — | |
| DMLPExtra data=Meta-knowledge, Self-supervised learning=true, Clean data percentage=50%2026.03 | 77.3 | — | — | |
| FasTEN w/o LCClean set usage=true, Loss Correction=false, Backbone=ResNet-50, Pre-trained=ImageNet2021.11 | 77.07 | — | — | |
| Self LearningExtra Data=true2023.05 | 76.44 | — | — | |
| AlexNetloss=cross-entropy, init=#1, training=1M, 50k2016.09 | 76.22 | — | — | |
| MSLGExtra Data=true2023.05 | 76.02 | — | — | |
| MSLG2022.02 | 76 | — | — | |
| MLC2022.02 | 75.8 | — | — | |
| MLaCClean set usage=true, Backbone=ResNet-50, Pre-trained=ImageNet2021.11 | 75.78 | — | — | |
| MLCExtra Data=true2023.05 | 75.78 | — | — | |
| LRA-diffusionConditioned on=CC* feature encoder2023.05 | 75.7 | — | — | |
| SANMbase_method=DivideMix2023.02 | 75.63 | — | — | |
| SANMExtra Data=false2023.05 | 75.63 | — | — | |
| SANM2023.05 | 75.63 | — | — | |
| DMLPExtra data=Meta-knowledge, Self-supervised learning=true, Clean data percentage=10%2026.03 | 75.5 | — | — | |
| CC2023.05 | 75.4 | — | — | |
| MFRW2022.11 | 75.35 | — | — | |
| CC*Backbone=ResNet50, Pre-processing=Sample selection via CC [44]2023.05 | 75.32 | — | — | |
| DivideMix+SNSCL2022.02 | 75.3 | — | — | |
| EPL2023.05 | 75.21 | — | — | |
| CLEANnoise structure=agnostic2018.05 | 75.2 | — | — | |
| Knockoffs-SPR2023.01 | 75.2 | — | — | |
| 50-ResNetloss=cross-entropy, init=ImageNet, training=50k2016.09 | 75.19 | — | — | |
| Nonedata=50k clean, pretrained=ImageNet2017.11 | 75.19 | — | — | |
| PGDF2021.12 | 75.19 | — | — | |
| VRIBackbone=ResNet-50, Pre-trained=ImageNet2026.03 | 75.19 | — | — | |
| DM-AugDescstrategy=WS-SAW2021.03 | 75.11 | — | — | |
| DM-AugDesc-WS-SAW2021.12 | 75.11 | — | — | |
| AugDescClean set usage=false, Backbone=ResNet-50, Pre-trained=ImageNet2021.11 | 75.11 | — | — | |
| AugDesc2023.02 | 75.11 | — | — | |
| AugDescExtra Data=false2023.05 | 75.11 | — | — | |
| AugDesc2022.11 | 75.11 | — | — | |
| AugDesc2022.02 | 75.1 | — | — | |
| Nested+Co-teachingBackbone=ResNet-50, Balanced subset or loss=true2022.06 | 75 | — | — | |
| PESEnsemble=true2021.06 | 74.99 | — | — | |
| UNICONBackbone=ResNet-502022.03 | 74.98 | — | — | |
| UNICON2023.05 | 74.98 | — | — | |
| WarPIBackbone=ResNet-50, Pre-trained=ImageNet2026.03 | 74.98 | — | — | |
| ProMixBackbone=ResNet-50, Pre-trained=ImageNet2022.07 | 74.94 | — | — | |
| CoDisBackbone=ResNet-50, Pre-trained=ImageNet2026.03 | 74.92 | — | — | |
| Nested-CoTeachingtraining_labels=noisy labels only2022.09 | 74.9 | — | — | |
| Nested2023.02 | 74.9 | — | — | |
| RRL2023.02 | 74.9 | — | — | |
| ULCBackbone=ResNet-50, Pre-trained=ImageNet2022.07 | 74.9 | — | — | |
| Nested2022.02 | 74.9 | — | — | |
| RRL2022.02 | 74.9 | — | — | |
| CLIPCleanerIntegrated module=DivideMix2024.08 | 74.87 | — | — | |
| MixFixBackbone=PreResNet-18, Zero-shot=false2024.08 | 74.87 | — | — | |
| C2D2024.08 | 74.84 | — | — | |
| C2DBackbone=PreResNet-18, Zero-shot=false2024.08 | 74.84 | — | — | |
| SSR+2024.08 | 74.83 | — | — | |
| kMEIDTM (+DivideMix)Integration=DivideMix, Implementation=Based on authors' code (*)2022.06 | 74.82 | — | — | |
| ELR+Backbone=ResNet-50, Pre-trained=ImageNet2020.06 | 74.81 | — | — | |
| ELR+2021.03 | 74.81 | — | — | |
| ELR+Initializer=ImageNet2021.04 | 74.81 | — | — | |
| ELR+Ensemble=true2021.06 | 74.81 | — | — | |
| ELR+2021.12 | 74.81 | — | — | |
| ELRBackbone=ResNet-502022.03 | 74.81 | — | — | |
| ELR+Backbone=ResNet-50, Pre-trained=ImageNet2022.07 | 74.81 | — | — | |
| ELR+Extra Data=false2023.05 | 74.81 | — | — | |
| ELR2023.05 | 74.81 | — | — | |
| ELR+2024.08 | 74.81 | — | — | |
| ELR+Backbone=ResNet-50, Pre-training=ImageNet, Mixup=true, Two Networks=true, Weight Averaging=true2021.08 | 74.81 | — | — | |
| ELRBackbone=ResNet-50, Pre-trained=ImageNet2026.03 | 74.81 | — | — | |
| ELR+Backbone=ResNet-50, Balanced subset or loss=true2022.06 | 74.8 | — | — | |
| DivideMixBackbone=ResNet-50, Balanced subset or loss=true2022.06 | 74.8 | — | — | |
| RRLBackbone=ResNet-502023.03 | 74.8 | — | — | |
| TCLBackbone=ResNet-502023.03 | 74.8 | — | — | |
| DivideMix2022.02 | 74.8 | — | — | |
| TCL2024.08 | 74.8 | — | — | |
| DivideMix2020.02 | 74.76 | — | — | |
| DivideMixBackbone=ResNet-50, Pre-trained=ImageNet2020.06 | 74.76 | — | — | |
| DivideMix2021.03 | 74.76 | — | — | |
| DivideMixInitializer=ImageNet2021.04 | 74.76 | — | — | |
| DivideMixBackbone=ResNet-50, Pre-trained=ImageNet2021.05 | 74.76 | — | — | |
| DivideMixEnsemble=true2021.06 | 74.76 | — | — | |
| DivideMix2021.12 | 74.76 | — | — | |
| DividemixSetting=CL2022.04 | 74.76 | — | — | |
| DMixBackbone=ResNet-502022.03 | 74.76 | — | — | |
| DivideMixClean set usage=false, Backbone=ResNet-50, Pre-trained=ImageNet2021.11 | 74.76 | — | — | |
| DivideMixPre-training=ImageNet2021.03 | 74.76 | — | — | |
| DivideMixtraining_labels=noisy labels only2022.09 | 74.76 | — | — | |
| DivideMix2021.03 | 74.76 | — | — |