Multi Deep Learning Model for Building Footprint Extraction from High Resolution Remote Sensing Image

EasyChair Preprint no. 7726

Abstract

3D city modeling is a new development trend in cartography that has a lot of practical and scientific value. The project necessitates the extraction of a building footprint using remote sensing images. This research examined how to solve the Building Footprint problem using automatic segmentation methods. To begin, we experiment with popular segmentation models such as Mask-RCNN, U-net, and U2-net. After that, we developed two multi-models that produced more stable and good results than the single models.

Keyphrases: building footprint, Convolution Neural Networks, remote sensing, Segmentation

@Booklet{EasyChair:7726,
  author = {Ho Trong Ánh and Tran Anh Tuan and Hoàng Phi Long and Lê Hai Hà and Tran Ngoc Thăng},
  title = {Multi Deep Learning Model for Building Footprint Extraction from High Resolution Remote Sensing Image},
  howpublished = {EasyChair Preprint no. 7726},

  year = {EasyChair, 2022}}