Table 6 Comparison of semantic segmentation model category accuracy.

From: Construction of a multiscale feature fusion model for indoor scene recognition and semantic segmentation

Category

U-Net ACC/%

DeeplabV3 + ACC/%

Seg-Net ACC/%

PSPNet ACC/%

Ours ACC/%

Wall

87.3

88.3

91.1

91.7

91.3

Floor

85.1

88.1

91.3

89.4

90.0

Ceiling

82.9

83.3

89.9

80.7

86.6

Bed

88.9

92.3

93.0

93.2

93.2

Windowpane

78.1

82.0

79.8

73.8

76.4

Cabinet

77.3

79.6

81.3

80.4

83.1

Person

75.0

71.3

80.8

Door

30.6

31.2

40.9

39.9

39.9

Table

52.7

63.9

74.4

67.0

69.9

Curtain

60.4

70.8

74.3

67.6

72.6

Chair

63.4

69.8

69.5

62.9

66.1

Sofa

42.7

53.1

Rug

18.6

45.7

49.2

48.0

46.8

Wardrobe

21.8

38.7

28.6

Light

22.2

59.7

50.5

56.5

62.9

Refrigerator

55.7

37.4

69.2

74.4

Pillow

44.9

39.9

39.1

43.1

Bookcase

48.2

58.7

59.7

Book

23.9

27.6

Computer

12.1

68.4

61.1

79.6

80.0

Cooktop

29.2

55.6

54.1

81.4

73.9

Ashcan

18.9

20.8

Fan

60.3

73.4

OA/%

75.9

79.5

80.1

80.9

82.6

  1. The ‘–’ symbol in the table indicates that the model either did not recognize the class or that the recognition accuracy was below 1%.