Comparative analysis of deep learning based building extraction methods with the new VHR Istanbul dataset

作者：

Highlights：

• A new building segmentation dataset for deep learning is generated.

• The current building datasets have been examined in terms of quality over quantity.

• Recent state-of-the-art architectures and encoders have been employed.

• Cross-evaluations with different datasets were performed for generability.

摘要

•A new building segmentation dataset for deep learning is generated.•The current building datasets have been examined in terms of quality over quantity.•Recent state-of-the-art architectures and encoders have been employed.•Cross-evaluations with different datasets were performed for generability.

论文关键词：Building extraction,Deep learning,Pléiades,Urban

论文评审过程：Received 12 March 2021, Revised 2 January 2022, Accepted 25 April 2022, Available online 2 May 2022, Version of Record 6 May 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.117346