Table 8 Performance on PASCAL VOC 2007 (vs. YOLOv11s: +0.4% \(\hbox {mAP}_{50}\), pottedplant +11.7%, tvmonitor +12.4%).

From: Multiscale diffusion-enhanced attention network for steel surface defect detection in Polysilicon Production

 

Classes

YOLOv5s

YOLOv7t

YOLOv8s

YOLOv10s

YOLOv11s

FFDDNet15

Ours

PASCAL VOC 2007

aeroplane

82.1

70.2

79.5

80

81.3

83.8

86.9

bicycle

89

79.4

80.9

86.3

88.7

89.8

90.8

bird

62.3

52.0

47.5

57.3

58.7

63.6

65.2

boat

76.2

49.9

58.3

74.3

79

64.6

75.2

bottle

63

40.3

50.4

63.3

66.3

70.8

60.5

bus

78.9

70.8

73.7

81.1

78.1

80.2

77.1

car

89.4

84.8

85.2

88.3

89

88.7

90.1

cat

83.6

65.1

74.7

86

82.2

79.1

84.8

chair

58.5

48.4

44.8

57

56.9

57.4

54

cow

73.9

66.1

57

78.7

78.2

75.3

78.4

diningtable

74.5

48.2

60.3

68.4

75.9

71.9

69.7

dog

74.3

61.9

59.9

73.7

76.4

74.8

79.9

horse

84.9

71.5

79

87.7

87.5

84.8

85

motorbike

91.6

76.8

87.5

89.9

93.4

89.6

93

person

87.1

79.9

82.6

85.7

88.1

86.7

87.6

pottedplant

37.6

41.7

27.5

43.2

41.8

46.0

46.7

sheep

57.6

63.2

45.3

51.5

59.7

60.7

56.1

sofa

64.9

60.0

47.6

65.6

68.8

69.4

64.6

train

86.1

77.6

79.5

85.3

87.1

88.1

89

tvmonitor

72.7

62.1

57.3

67.3

66.7

65.6

75.0

\(\hbox {mAP}_{50}\)

74.4

63.6

63.9

73.5

75.2

74.5

75.5