Duplex Restricted Network With Guided Upsampling for the Semantic Segmentation of Remotely Sensed Images
| Autor: | Longxue Liang, Xiaosuo Wu, Haowen Yan, Xiaoyu Wang, Jiali Cai, Wanzhen Lu |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
convolutional networks
General Computer Science Computer science Feature extraction 0211 other engineering and technologies Normalization (image processing) 02 engineering and technology Semantics Information distinction Upsampling remote sensing Encoding (memory) 0202 electrical engineering electronic engineering information engineering General Materials Science Segmentation Image resolution 021101 geological & geomatics engineering business.industry General Engineering Pattern recognition Image segmentation semantic segmentation TK1-9971 020201 artificial intelligence & image processing Artificial intelligence Electrical engineering. Electronics. Nuclear engineering business |
| Zdroj: | IEEE Access, Vol 9, Pp 42438-42448 (2021) |
| ISSN: | 2169-3536 |
| Popis: | Deep convolutional networks are of great significance for the automatic semantic annotation of remotely sensed images. Object position and semantic labeling are equally important in semantic segmentation tasks. However, the convolution and pooling operations of the convolutional network will affect the image resolution when extracting semantic information, which makes acquiring semantics and capturing positions contradictory. We design a duplex restricted network with guided upsampling. The detachable enhancement structure to separate opposing features on the same level. In this way, the network can adaptively choose how to trade-off classification and localization tasks. To optimize the detailed information obtained by encoding, a concentration-aware guided upsampling module is further introduced to replace the traditional upsampling operation for resolution restoration. We also add a content capture normalization module to enhance the features extracted in the encoding stage. Our approach uses fewer parameters and significantly outperforms previous results on two very high resolution (VHR) datasets: 84.81% (vs 82.42%) on the Potsdam dataset and 86.76% (vs 82.74%) on the Jiage dataset. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|