Table 3 Comparison with mainstream object detection models on the RSOD dataset.

From: Enhancing object detection in remote sensing images with improved YOLOv8 model

Model

mAP50-95 (%)

Para (M)

FPS

GFLOPs

YOLOv5n34

57.1

2.9

206.9

7.1

YOLOv8n28

62.1

3.3

223.7

8.1

RetinaNet35

31.0

19.8

25.0

153

EfficientDet36

50.5

18.4

19.5

102

Faster R-CNN17

62.1

41.4

26.4

157

TOOD37

63.2

32.0

21.8

147

Cascade R-CNN38

62.8

69.2

16.4

185

DETR20

46.2

32.6

20.3

71.3

Deformable DETR39

19.8

39.8

17.5

144

Ours

65.4

3.6

65.7

16.7

Back to article page