Object Detection on U.S. Golden dataset 300 train samples (test)
89.1mAP50SatlasNet
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SatlasNetTraining Strategy=Directly trained on Golden dataset2026.05 | 89.1 | 59 | 94.1 | 72.5 | 51.1 | |
| YOLO OursTraining Strategy=Pre-trained on auto-labeled dataset and fine-tuned on Golden training set2026.05 | 86.3 | 87.4 | 74.2 | 80.3 | 67.4 | |
| YOLO Golden +ECPTraining Strategy=Directly trained on Golden dataset, Hyperparameterization=ECP2026.05 | 80.2 | 71.4 | 74.4 | 72.9 | 50.8 | |
| F.Rcnn OursBackbone=Faster R-CNN, Training Strategy=Pre-trained on auto-labeled dataset and fine-tuned on Golden training set2026.05 | 77.5 | 65.7 | 79.8 | 72 | 48.1 | |
| F.Rcnn Golden +ECPBackbone=Faster R-CNN, Training Strategy=Directly trained on Golden dataset, Hyperparameterization=ECP2026.05 | 71.2 | 44.2 | 82.1 | 57.5 | 47.7 | |
| F.Rcnn AutoBackbone=Faster R-CNN, Training Strategy=Directly trained on auto-labeled dataset2026.05 | 60 | 19.3 | 86.9 | 31.6 | 39.5 | |
| YOLO AutoTraining Strategy=Directly trained on auto-labeled dataset2026.05 | 46.4 | 47.8 | 51.2 | 49.4 | 33.2 |