Table 5 Comparison experiment of semantic segmentation models.

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

Model

mIoU/%

OA/%

Precision/%

Recall/%

F1/%

Parameter/M

Time/s

U-Net

54.6

75.9

61.8

71.5

66.4

 ~ 33.4

 ~ 0.026

DeeplabV3 + 

59.7

79.5

66.3

77.9

71.7

 ~ 79.6

 ~ 0.055

Seg-Net

63.3

80.1

67.9

79.1

73.1

 ~ 55.1

 ~ 0.041

PSPNet

62.0

80.9

69.1

79.9

74.1

 ~ 90.6

 ~ 0.028

Ours

62.9

82.6

70.4

81.3

75.4

 ~ 360.0

 ~ 0.037

Back to article page