Coupling Mask R-CNN and Attention Mechanism for Building Extraction and Post-Processing Strategy

模型	mAP(Iou=0.5)	mAP(Iou=0.75)
Mask RCNN_Resnet101(原版)	0.577 3	0.425 7
Mask RCNN_RCBAnet101	0.620 4	0.479 3
Mask RCNN_Dsenet169	0.342 7	0.131 6
Mask RCNN_ $L_{m a s k - d i c e}$ (0.6,0.4)	0.619 3	0.438 9
Mask RCNN_ $L_{m a s k - d i c e}$ (0.7,0.3)	0.614 5	0.452 3
Mask RCNN_ $L_{m a s k - d i c e}$ (0.8,0.2)	0.580 0	0.422 9
Mask RCNN_RCBAnet101_ $L_{m a s k - d i c e}$ (0.7,0.3)	0.621 4	0.478 5
Mask RCNN_RCBAnet101_ $L_{m a s k - d i c e}$ (0.8,0.2)	0.315 5	0.109 1
Mask RCNN_RCBAnet101_ $L_{m a s k - d i c e}$ (0.6,0.4)	0.654 7	0.501 4

Table 2. Table 2There are four ways to segment the index value of Mask

View table

View in Article

Table 2. Table 2There are four ways to segment the index value of Mask

模型	召回率	精确度	F1-score
U-Net	0.698 4	0.633 3	0.664 2
Mask RCNN_Resnet101(原版)	0.693 5	0.691 3	0.692 3
Mask RCNN_RCBAnet101_ $L_{m a s k - d i c e}$ (0.6,0.4)	0.744 8	0.759 8	0.752 2
Mask RCNN_RCBAnet101_ $L_{m a s k - d i c e}$ (0.6,0.4)_后处理	0.744 0	0.761 0	0.752 4

Tools

Get Citation

Copy Citation Text

Buyu SU, Xiaoping DU, Haowei MU, Chen XU, Fang CHEN, Xiaonan LUO. Coupling Mask R-CNN and Attention Mechanism for Building Extraction and Post-Processing Strategy[J]. Remote Sensing Technology and Application, 2024, 39(3): 620

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites

Paper Information

Category:

Received: Nov. 9, 2022

Accepted: --

Published Online: Dec. 9, 2024

The Author Email:

DOI:10.11873/j.issn.1004-0323.2024.3.0620

Topics

laser devices and laser physics

Lasers and Laser Optics

Laser physics

laser manufacturing

Instrumentation, Measurement and Metrology