Search for HOT projects, news, people and jobs.

Integrating Deep Learning: AI-Assisted Humanitarian Mapping

HOT Program

Tanzania Uganda


HOT is partnering with Microsoft Philanthropies and Bing Maps to build a Machine Learning assisted workflow into the Tasking Manager application in order to enhance the project creation process and mapping experience.

The power of volunteer mapping is shown when disaster-affected areas can be mapped in a few days. When thousands of volunteers join together quickly, areas affected by a disaster then have the necessary data for a robust humanitarian response. Through our Missing Maps work, we continue to focus on preparedness mapping for unmapped areas but a challenge still exists to map at a scale and quality for humanitarian use. Around the world, there are still hundreds of millions of buildings and millions of kilometers of road missing from OpenStreetMap, and the data quality for humanitarians must be higher. To meet the challenge of mapping the unmapped areas, the Tasking Manager requires two key improvements to the workflow: an increase in data quality of edits being made, and an increase in mapping and validating efficiency.

Building upon the work the HOT community started in the fall of 2018, HOT is working with Microsoft’s AI for Humanitarian Action program to address these challenges by building enhanced assisted mapping workflows with machine learning into the Tasking Manager and create an open building dataset in Uganda and Tanzania.

Read about our notes on a machine learning direction.

There are promising potentials for machine/deep/statistical learning assisted mapping. Especially when it comes to image processing the achievements are tangible. HOT will pilot the use of this machine leveraged power of the newly available programming models to improve data quality and mapping experience. In this project, a well-designed use case around the Tasking Manager is to be implemented, which is focusing on machine learning assisted project creation and task facilitation.

Check out the machine learning primer from the World Bank.

We are teaming up with Microsoft, in order to provide an Open Source solution to include any kind of machine learning models into the Tasking Manager and other applications around OpenStreetMap. In addition through the project, Microsoft will be creating open building datasets for Uganda and Tanzania. These building datasets will be made available with an open license and used during the piloting of the new Tasking Manager workflows.

All this new work is thanks in part to the American Red Cross and the work they’ve done with Microsoft through the Missing Maps project with Microsoft. American Red Cross is a key partner in the project and will be helping work with us as we implement the work on the Tasking Manager.

Photo credit: OpenAerialMap

Recent news from Integrating Deep Learning: AI-Assisted Humanitarian Mapping (View all news)

Using machine learning to identify the effort and complexity of mapping areas

Testing more machine learning integrations with the Tasking Manager

Matthias Funke, Felix Delattre — 22 August, 2019

Hands on Assisted Tasks

Launch of an experimental playground Tasking Manager supported by state-of-the-art machine learning techniques to assist mappers, validators and project managers.

Felix Delattre — 29 May, 2019

User Experience Discovery with the Tasking Manager

Conducting design research to evaluate mapping and validation workflows within the Tasking Manager.

Leihla Pinho — 30 April, 2019