White blood cell classification on WBCBench 2026 (test)
0.742Macro F1 ScoreEns. (all)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Ens. (all)Backbones=ResNet50, ResNet152, Swin Transformer, Strategy=Probability averaging ensemble2026.03 | 0.742 | 0.771 | 0.726 | 0.994 | |
| Ens. (R50+R152)Backbones=ResNet50, ResNet152, Strategy=Probability averaging ensemble2026.03 | 0.724 | 0.727 | 0.73 | 0.994 | |
| Ens. (R152+Swin)Backbones=ResNet152, Swin Transformer, Strategy=Probability averaging ensemble2026.03 | 0.723 | 0.79 | 0.692 | 0.994 | |
| Ens. (R50+Swin)Backbones=ResNet50, Swin Transformer, Strategy=Probability averaging ensemble2026.03 | 0.72 | 0.791 | 0.69 | 0.994 | |
| SupervisedBackbone=Swin Transformer, Training Strategy=End-to-end supervised2026.03 | 0.712 | 0.69 | 0.764 | 0.994 | |
| DecoupledBackbone=ResNet152, Training Strategy=Two-stage decoupled learning2026.03 | 0.706 | 0.709 | 0.719 | 0.994 | |
| DecoupledBackbone=ResNet50, Training Strategy=Two-stage decoupled learning2026.03 | 0.704 | 0.72 | 0.7 | 0.993 | |
| SupervisedBackbone=ResNet152, Training Strategy=End-to-end supervised2026.03 | 0.687 | 0.661 | 0.755 | 0.994 | |
| DecoupledBackbone=Swin Transformer, Training Strategy=Two-stage decoupled learning2026.03 | 0.66 | 0.809 | 0.619 | 0.992 | |
| SupervisedBackbone=ResNet50, Training Strategy=End-to-end supervised2026.03 | 0.658 | 0.632 | 0.712 | 0.993 |