eduzhai > Applied Sciences > Engineering >

MACU-Net for Semantic Segmentation of Fine-Resolution Remotely Sensed Images

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Semantic segmentation of remotely sensed images plays an important role inland resource management, yield estimation, and economic assessment. U-Net, adeep encoder-decoder architecture, has been used frequently for imagesegmentation with high accuracy. In this Letter, we incorporate multi-scalefeatures generated by different layers of U-Net and design a multi-scale skipconnected and asymmetric-convolution-based U-Net (MACU-Net), for segmentationusing fine-resolution remotely sensed images. Our design has the followingadvantages: (1) The multi-scale skip connections combine and realign semanticfeatures contained in both low-level and high-level feature maps; (2) theasymmetric convolution block strengthens the feature representation and featureextraction capability of a standard convolution layer. Experiments conducted ontwo remotely sensed datasets captured by different satellite sensorsdemonstrate that the proposed MACU-Net transcends the U-Net, U-NetPPL, U-Net3+, amongst other benchmark approaches. Code is available atthis https URL.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×