Residential building facade segmentation in the urban environment

Building retrofit is an important facet in the drive to reduce global greenhouse gas emissions. However, delivering building at scale a significant challenge, especially how automate process of surveying. On-site survey by expert surveyors main approach industry. This can lead high workload if planning large-scale. An advanced vehicle-mounted data capturing system has been built collect urban environmental multi-spectral data. The contains substantial information that essential identifying needs. Although able highly-efficient manner, analysis still big challenge apply into plans. In this paper, street-view facade image segmentation model designed as foundation holistic framework. developed on deep learning-based semantic technology and uses ensemble learning strategy. object detection fused magnifier improve performance small objects boundary predictions. achieved state-of-the-art levels accuracy dataset.