Semantic Segmentation of Remote Sensing Image Based on Convolutional Neural Network and Mask Generation
Table 3
Evaluation indicators of each method on the Vaihingen dataset (%).
Category
Evaluation
Low vegetation
Buildings
Impervious surfaces
Cars
Trees
U-Net
Precision
69.6
60.2
78.2
70.2
78.8
Recall
79.7
71.5
79.7
76.2
82.2
IoU
58.7
62.6
71.7
65.4
68.9
CASIA3
Precision
81.8
71.8
79.3
73.5
81.8
Recall
76.6
82.6
80.6
85.7
82.6
IoU
72.2
64.6
69.9
64.8
78.8
HUSTW5
Precision
86.3
74.3
81.5
79.7
70.8
Recall
81.2
86.9
85.6
86.5
86.6
IoU
71.2
67.3
79.7
72.5
79.3
Proposed
Precision
86.9
83.4
89.3
82.7
89.9
Recall
87.2
87.3
87.4
88.2
91.4
IoU
73.2
70.3
81.2
72.9
82.7
The bold values represent the maximum values of evaluation index in the same column, such as precision, recall, and IoU and the index values of this paper are greater than those of other algorithms.