Table 5 Comparison of different FPN architectures.

From: LRDS-YOLO enhances small object detection in UAV aerial images with a lightweight and efficient design

Model

Precision (%)

Recall (%)

mAP50 (%)

mAP50:95 (%)

Parameters (M)

GFlops

PANet

46.3

35.3

35.7

21

26.7

10.4

BiFPN

48.2

36.9

37.9

22.4

21

19.5

AFPN

47.9

37.7

38.3

22.9

29.0

19.6

Re-Calibration FPN

49.9

34.0

39.5

24.2

3.85

18.8

Back to article page