Object Detection on VisDrone (test-dev)
61.3mAP@0.5Full Fine-tuning
Evaluation Results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Full Fine-tuning∆ Params.=0, ∆ Trainable.=27.5M2026.07 | 61.3 | — | — | — | 38.4 | — | — | — | — | — | — | — | |
| DroneFINE-T (Ours)∆ Params.=1.54M, ∆ Trainable.=1.54M2026.07 | 61.2 | — | — | — | 38.4 | — | — | — | — | — | — | — | |
| Mona∆ Params.=1.4M, ∆ Trainable.=1.4M2026.07 | 60.5 | — | — | — | 37.9 | — | — | — | — | — | — | — | |
| Adapter∆ Params.=1.14M, ∆ Trainable.=1.14M2026.07 | 60.1 | — | — | — | 37.4 | — | — | — | — | — | — | — | |
| Shine∆ Params.=0.02M, ∆ Trainable.=0.02M2026.07 | 60.1 | — | — | — | 37 | — | — | — | — | — | — | — | |
| CoOp∆ Params.=0.012M, ∆ Trainable.=0.012M2026.07 | 60 | — | — | — | 37.3 | — | — | — | — | — | — | — | |
| Baseline∆ Params.=0, ∆ Trainable.=02026.07 | 59.6 | — | — | — | 37.1 | — | — | — | — | — | — | — | |
| NormTuning∆ Params.=0, ∆ Trainable.=0.025M2026.07 | 59.6 | — | — | — | 37.1 | — | — | — | — | — | — | — | |
| VPT∆ Params.=0.08M, ∆ Trainable.=0.08M2026.07 | 59.6 | — | — | — | 37.1 | — | — | — | — | — | — | — | |
| LoRA∆ Params.=1.13M, ∆ Trainable.=1.13M2026.07 | 59.3 | — | — | — | 36.7 | — | — | — | — | — | — | — | |
| Proposed ModelVision Backbone=Proposed YOLOv8, Text Model=CLIP, GFLOPs=15.2, Params (M)=3.82026.02 | 30.7 | 41.6 | 31 | 35.5 | — | — | — | — | — | — | — | — | |
| YOLO-WorldVision Backbone=YOLOv8, Text Model=CLIP, GFLOPs=15.7, Params (M)=42026.02 | 30.4 | 40.6 | 30.8 | 35 | — | — | — | — | — | — | — | — | |
| YOLOv9Vision Backbone=YOLOv9, GFLOPs=7.9, Params (M)=22026.02 | 30.2 | 42.3 | 30.2 | 35.2 | — | — | — | — | — | — | — | — | |
| YOLOv11Vision Backbone=YOLOv11, GFLOPs=6.3, Params (M)=2.52026.02 | 29.8 | 40.2 | 30.6 | 34.7 | — | — | — | — | — | — | — | — | |
| YOLOv10Vision Backbone=YOLOv10, GFLOPs=8.2, Params (M)=2.62026.02 | 29.7 | 40.4 | 30.3 | 34.6 | — | — | — | — | — | — | — | — | |
| Zero-shot Detection YOLOVision Backbone=YOLOv5, Text Model=CLIP, GFLOPs=18.5, Params (M)=7.32026.02 | 29.1 | 39.3 | 29.9 | 33.9 | — | — | — | — | — | — | — | — | |
| Zero-shot∆ Params.=0, ∆ Trainable.=–2026.07 | 26.9 | — | — | — | 16.5 | — | — | — | — | — | — | — | |
| DLWBackbone=EfficientDet2026.02 | 12.1 | — | — | — | 6.5 | 1.1 | 10 | 33.2 | 9.7 | 2.9 | 15.7 | 42.1 | |
| DLSBackbone=EfficientDet2026.02 | 11.9 | — | — | — | 6.6 | 1 | 10 | 33.3 | 9.7 | 2.8 | 15.8 | 40.8 | |
| BaselineBackbone=EfficientDet2026.02 | 11.2 | — | — | — | 6.2 | 0.9 | 9.5 | 33.4 | 9.3 | 2.7 | 15.2 | 40.8 | |
| BWBackbone=EfficientDet2026.02 | 11 | — | — | — | 6.1 | 0.9 | 9.1 | 35.7 | 9.1 | 2.5 | 14.7 | 45.4 | |
| IWBackbone=EfficientDet2026.02 | 11 | — | — | — | 6.1 | 0.9 | 9.3 | 34.2 | 9.2 | 2.6 | 14.8 | 43.3 | |
| CLBackbone=EfficientDet2026.02 | 6.8 | — | — | — | 3.7 | 0.6 | 5.1 | 21.8 | 6 | 1.7 | 9.2 | 30.8 | |
| LBackbone=EfficientDet2026.02 | 5.4 | — | — | — | 3.1 | 0.4 | 3.9 | 19.4 | 4.7 | 1.4 | 6.3 | 24.3 |