Image Classification on ImageNet 10% labels 1k (train)
91.9Top-5 AccCustomKD
Evaluation results
| Method | Links | ||
|---|---|---|---|
| CustomKDLabel Ratio=10%2025.03 | 91.9 | 74.7 | |
| SimMatchLabel Ratio=10%2025.03 | 91.6 | 74.4 | |
| CoMatchLabel Ratio=10%2025.03 | 91.4 | 73.7 | |
| OBoWEpochs=200, Batch=256, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 90.7 | — | |
| SwAVEpochs=800, Batch=4096, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 89.9 | — | |
| FixMatchLabel Ratio=10%2025.03 | 89.1 | 71.5 | |
| BYOLEpochs=1000, Batch=4096, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 89 | — | |
| SwAVEpochs=200, Batch=256, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 88.7 | — | |
| Vat + EntMinLabel Ratio=10%2025.03 | 88.5 | 68.8 | |
| UDALabel Ratio=10%2025.03 | 88.5 | 68.8 | |
| SimCLREpochs=1000, Batch=4096, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 87.8 | — | |
| PCLEpochs=200, Batch=256, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 85.6 | — | |
| PIRLEpochs=800, Batch=1024, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 83.8 | — | |
| S4L-RotationLabel Ratio=10%2025.03 | 83.8 | 53.4 | |
| SupervisedEpochs=100, Batch=256, Evaluation Protocol=Fine-tuning, Backbone=ResNet-50, Testing Mode=single-crop testing2020.12 | 80.4 | — | |
| ResNet50Backbone=ResNet-502021.12 | 50.86 | — | |
| Ours-JointTraining strategy=Joint (without weight sharing)2021.12 | 49.73 | — | |
| Agent-BModel=Agent-B2021.12 | 47.46 | — | |
| Ours-SharedTraining strategy=Shared (with weight sharing)2021.12 | 45.75 | — | |
| ViT-B-SparseBackbone=ViT-B, Variant=Sparse2021.12 | 28.92 | — | |
| ViT-B-SAMBackbone=ViT-B, Optimizer=SAM2021.12 | 28.66 | — | |
| ViT-BBackbone=ViT-B2021.12 | 28.11 | — |