It is a cutting-edge machine learning model that harnesses high-resolution satellite imagery to swiftly analyze and identify urban buildings, accomplishing this task up to 8 times faster than manual methods.
This innovative detector employs a semantic segmentation approach to predict raster masks and vectorized polygons from satellite images. It bestows each pixel with a categorical assignment, enabling its versatility across diverse contexts by recognizing the distinctive traits of each city. The model's training encompassed imagery from Georgetown, Paramaribo, and Belize, refining its capacity to perform accurately.
The overarching goal of this model is to furnish urban planners with an automated tool for generating fundamental maps of existing buildings within urban landscapes and human settlements. By significantly expediting the identification process, it empowers planners to efficiently access essential data, enhancing their ability to strategize and make informed decisions concerning urban development.