Table 5 Performance comparison between MFRA-YOLO AND YOLOv8n in each category on the VISDRONE2019 dataset.

From: An improved UAV image object detection algorithm combining multi-scale feature fusion and receptive-field attention-based convolution

Categories

Validation

Test

mAP50%

mAP50:95%

mAP50%

mAP50:95%

YOLOv8n

MFRA-YOLO

YOLOv8n

MFRA-YOLO

YOLOv8n

MFRA-YOLO

YOLOv8n

MFRA-YOLO

Pedestrian

36

39.7

15.6

17.7

22.4

25.8

8.82

10.3

People

28.3

31.7

10.5

11.7

12.5

13.7

4.2

4.76

Bicycle

8.6

9.62

3.21

3.91

5.24

7.74

2.11

3.05

Car

76.3

78.6

53

55.3

66.5

70.5

40.8

43.6

Van

39.3

43.4

26.9

30.3

30.9

35.9

19.7

22.6

Truck

30

34.2

19.8

22.2

29.6

34.7

17.9

21.4

Tricycle

22.2

26.4

12

14.2

13.3

15.9

6.66

8.1

Awning-Tricycle

12.3

13.9

7.75

8.78

14.5

18.5

7.73

9.67

Bus

49.3

52.3

34.1

37.3

49.9

55.2

32.7

37.2

Motor

37.4

40.8

15.7

17.7

23.1

26.1

8.92

10.2

Back to article page