Active Fine-tuning on TinyImageNet
30F-acc (2%)BRAL-T
Evaluation results
| Method | Links | ||
|---|---|---|---|
| BRAL-TBackbone=Deit-Small, Pre-trained=DINO framework on ImageNet-1k, Resolution=224 x 224, Initial labeled subset=1%, Active learning iteration increment=1%, Batch size=512, Epochs=1000, Optimizer=SGD2026.04 | 30 | 39.2 | |
| ActiveFTBackbone=Deit-Small, Pre-trained=DINO framework on ImageNet-1k, Resolution=224 x 224, Initial labeled subset=1%, Active learning iteration increment=1%, Batch size=512, Epochs=1000, Optimizer=SGD2026.04 | 28.9 | 35.9 | |
| RandomSampleBackbone=Deit-Small, Pre-trained=DINO framework on ImageNet-1k, Resolution=224 x 224, Initial labeled subset=1%, Active learning iteration increment=1%, Batch size=512, Epochs=1000, Optimizer=SGD2026.04 | 21.3 | 34.8 |