Object Detection on MAVREC aerial (test)
61.03mAPFlorence-2 + CPA + PCL
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Florence-2 + CPA + PCLBackbone=Florence-2-base (230M parameters)2026.05 | 61.03 | 73.13 | 66.91 | 60.44 | 83.24 | |
| Florence-2 + Complexity-Aware Pathway AggregationBackbone=Florence-2-base (230M parameters)2026.05 | 60.66 | 72.61 | 66.48 | 59.51 | 85.69 | |
| Florence-2 (random)Backbone=Florence-2-base (230M parameters)2026.05 | 58.66 | 71.06 | 64.19 | 59.67 | 79.81 | |
| Florence-2 + Paired Curriculum LearningBackbone=Florence-2-base (230M parameters)2026.05 | 56.53 | 68.4 | 61.88 | 56.2 | 81.2 | |
| D-DETRPre-trained on COCO=true2026.05 | 33.2 | 61.9 | — | 31.5 | 51 | |
| YOLOv72026.05 | 31.9 | 58.8 | — | 31.4 | 63.1 | |
| YOLO-NAS-L2026.05 | 27 | 43.3 | — | 25.9 | 58 | |
| DETR2026.05 | 23.6 | 40.1 | — | 23.4 | 44.9 | |
| D-DETRPre-trained on COCO=false2026.05 | 10.3 | 25 | — | 10.1 | 29.4 |