Aiding Humanitarian Workers With Technology

Whether it's for disaster recovery, risk reduction or educational programmes, humanitarian workers can’t meet the needs of vulnerable communities unless they know where those communities are. That is the goal of Missing Maps, an open, collaborative fieldwork project in which volunteers map vulnerable places by tracing dwellings and other features onto satellite images. The resulting maps are open source, and available in OpenStreetMap.

It’s a time-consuming process. The collected data needs to be checked before volunteers on the ground add details such as local street and road names, and humanitarian centres. At present, many projects do not get completed fast enough, leaving humanitarian teams without the detailed maps they need.

Aiding the aid workers

A team from Occam DM – the newly formed Occam Watson Guild – was looking for a new project to harness the cognitive capability of IBM Watson and enter the Watson Build, IBM’s first cognitive challenge designed solely for Business Partners. Automating identification and classification of dwellings using satellite imagery provided by the Missing Maps project proved the ideal challenge. Once complete, the project should enable faster completion of more detailed maps, saving aid workers time in planning and executing their projects, logistics, and ensuring they have visibility of all dwellings and homesteads once they arrive.

The system could also aid disaster response efforts, identifying areas of greatest damage and helping direct relief efforts.

The custom classification challenge

If this project is to be successful, the existing Watson Visual Recognition software needs to be trained to identify and classify dwellings in the African bush. This presents particular challenges for the Guild team for two main reasons:

i) Satellite images can vary widely in quality and colour

ii) There is a huge degree of variability in terms of dwelling styles, size and construction materials

Testing the concept and making a difference

To overcome these challenges and test the accuracy of the classifier, the project team has focused on one area in Tanzania. The area has been chosen for its relative consistency in terms of dwellings and ground cover, but also because it is a location where the project team can make a real, immediate difference. December marks the start of FGM (female genital mutilation) cutting season in the area. Developing improved maps for this area now could aid organisations working in the area to plan educational visits to villages and homesteads.

As they are developed and trained, each custom classifier is run against areas that are simultaneously manually classified, to ensure accuracy.

Increasing coverage:

Once the team is satisfied with results at the test site, they will extend the classifier’s capabilities, increasing the pool of dwelling types, identifying access paths and animal enclosures, and expanding geographical coverage.

Simon Peel, Technical Consultant at Occam DM says: “The Occam Watson Guild was formed to push the boundaries of our adopted IBM technology stack and to explore new and complementary technologies. We are delighted to be doing that in way that can make a real difference to vulnerable communities.”

The team’s project has now been selected as a European finalist. It now progresses to build stage, with a prototype due to be presented to a panel of judges in early October.

Neth Williams- Marketing Manager, Occam DM Ltd (part of the St Ives Group)