Detection of informal settlements using optical and SAR satellite images

5 HPC Summer School Official Website
https://cybercolombia.org/summer_scho...

Speaker: María Roberta Devesa (Dymaxion Labs)

Date/Time (COT): 10:30 - 11:30

Abstract: With the lack of urban planning policies, the generation of informal settlements grows up quickly. These settlements are usually in low-quality areas, many times with no resources like clean water or electricity. And on several occasions, they are located in vulnerable areas with flood risk or landslides risk. Dus, it is essential for the governments to know locations to include them in the urban planning and avoid catastrophes during storms or other sources. The analysis goal is to detect informal settlements using satellite images and deep neural networks to process big extensions of areas fast and at a low cost.

To address this topic, two different techniques are presented. The former uses optical images from the Sentinel-2 satellite. The second uses radar images from the Sentinel-1 satellite. But, both process the images using a deep convolutional neural network with a U-Net architecture. The advantages and limitations of each study are presented and compared for the given results.

Bio: M. Roberta Devesa is a Ph.D. in Physics, who worked at CERN from 2014 to 2019 doing studies on new physics and statistical analysis. Currently, since 2020, she works at DymaxionLabs as a Sr. data scientist. Developing in computer vision and research for several projects.