Object Detection on UAVDT (test)
76.55mAPYolov4 with PRB-FPN
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Yolov4 with PRB-FPNBackbone=CSPDarknet-53, input size=512x512, FPS=19.22020.12 | 76.55 | — | — | — | — | — | — | — | |
| PRB-FPNBackbone=MobileNet-V2, input size=512x512, FPS=752020.12 | 65.47 | — | — | — | — | — | — | — | |
| Yolov4 with Re-COREBackbone=CSPDarknet-53, input size=512x512, FPS=26.32020.12 | 65.41 | — | — | — | — | — | — | — | |
| Yolov4 with BFMBackbone=CSPDarknet-53, input size=512x512, FPS=30.52020.12 | 64.52 | — | — | — | — | — | — | — | |
| SpotNetBackbone=Hourglass-104, input size=512x5122020.12 | 52.8 | — | — | — | — | — | — | — | |
| CenterNetBackbone=Hourglass-104, input size=512x5122020.12 | 51.18 | — | — | — | — | — | — | — | |
| LRFNetBackbone=VGG-16, input size=512x512, FPS=912020.12 | 37.81 | — | — | — | — | — | — | — | |
| R-FCNBackbone=ResNet-50, input size=1024x540, FPS=4.72020.12 | 34.35 | — | — | — | — | — | — | — | |
| Re-COREBackbone=MobileNet-V2, input size=512x512, FPS=1102020.12 | 34.2 | — | — | — | — | — | — | — | |
| RetinaNetBackbone=ResNet-101-FPN, input size=512x512, FPS=252020.12 | 33.95 | — | — | — | — | — | — | — | |
| SSDBackbone=VGG-16, input size=512x512, FPS=41.562020.12 | 33.62 | — | — | — | — | — | — | — | |
| BFMBackbone=MobileNet-V2, input size=512x512, FPS=1132020.12 | 29.7 | — | — | — | — | — | — | — | |
| DroneFINE-T (Ours)∆ Params.=1.54M, ∆ Trainable.=1.54M2026.07 | 26.5 | 43.6 | — | — | — | — | — | — | |
| DroneFINE-T2026.07 | 26.5 | 43.6 | — | — | — | — | — | — | |
| LoRA∆ Params.=1.13M, ∆ Trainable.=1.13M2026.07 | 25 | 40.8 | — | — | — | — | — | — | |
| CoOp∆ Params.=0.012M, ∆ Trainable.=0.012M2026.07 | 24.9 | 41.4 | — | — | — | — | — | — | |
| Baseline∆ Params.=0, ∆ Trainable.=02026.07 | 24.7 | 40.9 | — | — | — | — | — | — | |
| UFPMP-Det2026.07 | 24.6 | 38.7 | — | — | — | — | — | — | |
| Shine∆ Params.=0.02M, ∆ Trainable.=0.02M2026.07 | 24.3 | 42.4 | — | — | — | — | — | — | |
| ESOD2026.07 | 23.6 | 47.6 | — | — | — | — | — | — | |
| NormTuning∆ Params.=0, ∆ Trainable.=0.025M2026.07 | 22.7 | 40.1 | — | — | — | — | — | — | |
| Adapter∆ Params.=1.14M, ∆ Trainable.=1.14M2026.07 | 22.5 | 39.3 | — | — | — | — | — | — | |
| Full Fine-tuning∆ Params.=0, ∆ Trainable.=27.5M2026.07 | 22.4 | 38.6 | — | — | — | — | — | — | |
| Faster-RCNNBackbone=VGG-16, input size=1024x540, FPS=2.82020.12 | 22.32 | — | — | — | — | — | — | — | |
| VPT∆ Params.=0.08M, ∆ Trainable.=0.08M2026.07 | 22.3 | 37.8 | — | — | — | — | — | — | |
| Mona∆ Params.=1.4M, ∆ Trainable.=1.4M2026.07 | 22.2 | 37.6 | — | — | — | — | — | — | |
| FRFDet-LVariant=L, Input Size=640x6402026.07 | 21.7 | 32.7 | 25 | — | — | — | — | — | |
| RON**Backbone=VGG-16, input size=512x512, FPS=11.12020.12 | 21.59 | — | — | — | — | — | — | — | |
| SFFNet-X2026.04 | 20.6 | 34.4 | 21.9 | — | — | 14.3 | 31.5 | 33.3 | |
| SFFNet-Xsize=X2026.04 | 20.6 | 34.4 | 21.9 | — | — | 14.3 | 31.5 | 33.3 | |
| RemDet-L2026.07 | 20.6 | 34.5 | — | — | — | — | — | — | |
| RemDet-LVariant=L, Input Size=640x6402026.07 | 20.6 | 34.5 | 22.1 | — | — | — | — | — | |
| SCLNet2026.04 | 20 | 33.1 | 22.3 | — | — | — | — | — | |
| SCLNet2026.04 | 20 | 33.1 | 22.3 | — | — | — | — | — | |
| YOLOv10-Xreimplemented=true2026.04 | 19.5 | 32.3 | 21 | — | — | 13.4 | 28.9 | 31.5 | |
| YOLOv10-Xreimplementation=true, size=X2026.04 | 19.5 | 32.3 | 21 | — | — | 13.4 | 28.9 | 31.5 | |
| YOLCInput Size=640x6402026.07 | 19.3 | 30.9 | 20.1 | — | — | — | — | — | |
| PRDET2026.04 | 19.1 | 33.8 | 19.8 | — | — | — | — | — | |
| PRDET2026.04 | 19.1 | 33.8 | 19.8 | — | — | — | — | — | |
| GLSAN2026.04 | 19 | 30.5 | 21.7 | — | — | — | — | — | |
| YOLO11-Xreimplemented=true2026.04 | 19 | 33 | 19.6 | — | — | 13 | 30.7 | 32.2 | |
| GLSAN2026.04 | 19 | 30.5 | 21.7 | — | — | — | — | — | |
| YOLOv11-Xreimplementation=true, size=X2026.04 | 19 | 33 | 19.6 | — | — | 13 | 30.7 | 32.2 | |
| GLSDetInput Size=640x6402026.07 | 18.3 | 29.8 | 17.6 | — | — | — | — | — | |
| AMRNet2026.04 | 18.2 | 30.4 | 19.8 | — | — | 10.3 | 31.3 | 33.5 | |
| AMRNet2026.04 | 18.2 | 30.4 | 19.8 | — | — | 10.3 | 31.3 | 33.5 | |
| AMRNetInput Size=640x6402026.07 | 18.2 | 30.4 | 19.8 | — | — | — | — | — | |
| YOLOv8-Xreimplemented=true2026.04 | 18.1 | 31.1 | 18.8 | — | — | 13 | 28.4 | 32.4 | |
| YOLOv8-Xreimplementation=true, size=X2026.04 | 18.1 | 31.1 | 18.8 | — | — | 13 | 28.4 | 32.4 | |
| EFCInput Size=640x6402026.07 | 18 | 31.5 | 18.9 | — | — | — | — | — | |
| CEASCBase Detector=GFL V12023.03 | 17.1 | 30.9 | 17.8 | 64.12 | 28.47 | — | — | — | |
| CEASC2026.04 | 17.1 | 30.9 | 17.8 | — | — | — | — | — | |
| CEASC2026.04 | 17.1 | 30.9 | 17.8 | — | — | — | — | — | |
| CEASC2026.07 | 17.1 | 30.9 | — | — | — | — | — | — | |
| CEASCInput Size=640x6402026.07 | 17.1 | 30.9 | 17.8 | — | — | — | — | — | |
| GFLreimplemented=true2026.04 | 17 | 30.3 | 18.1 | — | — | 11.2 | 28.5 | 30.9 | |
| GFLreimplementation=true2026.04 | 17 | 30.3 | 18.1 | — | — | 11.2 | 28.5 | 30.9 | |
| GLSANInput Size=640x6402026.07 | 17 | 28.1 | 18.8 | — | — | — | — | — | |
| GFL V1 (Baseline)Base Detector=GFL V12023.03 | 16.9 | 29.5 | 17.9 | 271.66 | 20.49 | — | — | — | |
| GFL V1Input Size=640x6402026.07 | 16.9 | 29.5 | 17.9 | — | — | — | — | — | |
| CDMNetInput Size=640x6402026.07 | 16.8 | 29.1 | 18.5 | — | — | — | — | — | |
| DMNet2026.04 | 14.7 | 24.6 | 16.3 | — | — | 9.3 | 26.2 | 35.2 | |
| DMNet2026.04 | 14.7 | 24.6 | 16.3 | — | — | 9.3 | 26.2 | 35.2 | |
| DMNetInput Size=640x6402026.07 | 14.7 | 24.6 | 12.5 | — | — | — | — | — | |
| ClusDet2026.04 | 13.7 | 26.5 | 12.5 | — | — | 9.1 | 25.1 | 31.2 | |
| ClusDet2026.04 | 13.7 | 26.5 | 12.5 | — | — | 9.1 | 25.1 | 31.2 | |
| ClusDet2026.07 | 13.7 | 26.5 | — | — | — | — | — | — | |
| ClusDetInput Size=640x6402026.07 | 13.7 | 26.5 | 12.5 | — | — | — | — | — | |
| CenterNet2026.04 | 13.2 | 26.7 | 11.8 | — | — | 7.8 | 26.6 | 13.9 | |
| CenterNet2026.04 | 13.2 | 26.7 | 11.8 | — | — | 7.8 | 26.6 | 13.9 | |
| CenterNetInput Size=640x6402026.07 | 13.2 | 26.7 | 11.8 | — | — | — | — | — | |
| Faster-RCNNInput Size=640x6402026.07 | 11 | 23.4 | 8.4 | — | — | — | — | — | |
| Zero-shot∆ Params.=0, ∆ Trainable.=–2026.07 | 9.5 | 18.5 | — | — | — | — | — | — | |
| LAE-DINO2026.07 | — | 36.5 | — | — | — | — | — | — | |
| RT-OVAD2026.07 | — | 40.5 | — | — | — | — | — | — | |
| YOLO-World2026.07 | — | 35.8 | — | — | — | — | — | — |