Object Detection on SPARK 2024 (test)
99.2AP (%)Oracle
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OracleModel=YOLOv11, Training=Ground-truth annotations2026.02 | 99.2 | 99.5 | 99.5 | |
| OracleModel=Efficient-Det, Training=Ground-truth annotations2026.02 | 98.7 | 99 | 99 | |
| TTA+WBF+CF+TRModel=YOLOv11, Protocol=Train Relabeling / Iterative Distillation2026.02 | 76.4 | 99.5 | 98.2 | |
| TTA+WBF+CFModel=YOLOv11, Protocol=Confidence Filtering (0.5)2026.02 | 70.8 | 99.5 | 97.4 | |
| TTA+WBFModel=YOLOv11, Protocol=Pseudo-labels distillation2026.02 | 68.9 | 99.5 | 89.4 | |
| TTA+WBF+CF+TRModel=Efficient-Det, Protocol=Train Relabeling / Iterative Distillation2026.02 | 68.7 | 98.9 | 80.5 | |
| TTA+WBF+CFModel=Efficient-Det, Protocol=Confidence Filtering (0.5)2026.02 | 66.8 | 98.9 | 80.4 | |
| TTA+WBFModel=Efficient-Det, Protocol=Pseudo-labels distillation2026.02 | 65 | 98.9 | 78.9 |