Kamal-Eldin/UNICEF-Giga_Sudan

Detecting school lots from satellite imagery in Southern Sudan using a UNET segmentation model. This project is done to assist the UNICEF's Giga initiative in their mission to connect unconnected schools worldwide to the internet. # UNICEF-Giga_Sudan ## INTRODUCTION This repo shows my contribution as an ML engineer to the UNICEF-Giga project organized by Omdena to help connect schools in developing nations worldwide to the internet. For a more detailed walkthrough of the technical solution please click the colab icon below: or download the notebook in this repo at Walkthrough_Model_Development.ipynb ### Giga The UNICEF Office of Innovation describes Giga as: > Giga, an initiative launched by UNICEF and ITU in September 2019 to connect every school to the Internet and every young person to information, opportunity and choice, is supporting the response to COVID-19, as well as looking at how connectivity can create stronger infrastructures of hope and opportunity in the "time after COVID." Giga's effort is mainly focused on developing nations, where its often the case that school records are either noncomplete or entirely nonexistent, with a lot of non-official locally sourced schools. Giga's team could benefit from an automated school discovery and localization solution to assit their mission with speed and efficiency, while reducing the costs and the often scarce manpower involved in ground surveys. Omdena is an "AI for Good" company that summons voluneering efforts worldwide to develop and deploy AI tools for positive impact. Omdena organized this volunteer effort to develop an AI solution for the UNICEF to support Giga's mission. I had the honour of making a contribution to such endeavor, shown here in this repo. ## PROBLEM STATMENT Giga is resorting to satellite imagery instead of ground surveys to locate schools worldwide. Satellite imagery provides wide coverage and speed. Giga is developing AI systems that utilize these satellite data to automate the detection of schools, with a focus on higher recall rates, but also at an adequate precision. ***A recall first*** principle helps Giga to discover and include as much schools as possible, while maintaining an adequate precisio …