How to Automate the Analysis of Aerial Data in the Context of Aid & Development

WeRobotics, Picterra, Pix4D, Geneva Science-Policy Interface, International Organization for Migration (IOM), Medair, UNOSAT, UNITAR, ETH Zürich Ecovision, EPFL Tech4Impact

Context

On 11 April, at the 2019 WSIS Forum hosted by the ITU, the Geneva Science-Policy Interface, in partnership with WeRobotics and other partners, organized a panel discussion on the use of aerial data in the context of aid and development. This event was a direct follow-up to an expert meeting that the GSPI hosted in December 2018.

The panel shed light on a type of science-policy-engagement model that GSPI is promoting, with very concrete examples of how IOs, NGOs and scientists can work together to bring about impactful, evidence-based solutions to current challenges.

In particular, this session focused on how aerial images produced by drones or satellites can provide vital data to support the work of aid & development workers. For instance, such data can help the monitoring of refugee settlements, emergency response to natural disasters, scalable agriculture programmes, and much more.

While data has become more available to support evidence-based decision making, today the challenge has shifted from ‘no data’ to ‘too much data’, thus calling for automated solutions.

Summary of discussion

“An image contains a lot of information, and an AI is a way to go through all of it and retrieve useful information.”

The discussion covered four overarching topics: the current needs of organizations in terms of data analysis; existing methods; concrete applications; and coordination of efforts.

Organizations need to make sense of large databases to understand complex, multi-faceted issues and act quickly. This effort raises many analytical challenges as data analysis still is very resource-intensive for many organizations. For instance, MEDAIR’s current bottlenecks are the slow pace of data collection and analysis; it is not done systematically; and the difficulties in training their staff to work in data science. Additionally, refugee camps are crowded and dense. Environmental issues are large in scale and interrelated. Conflict-affected zones are difficult to access. This clouds organizations’ sight and ability to design appropriate programmes. Therefore, how can organizations see more clearly to intervene more effectively?

Several panellists presented methods which directly respond to organizations’ needs. PIX4D and Picterra presented the state-of-the-art means of bridging remote-sensing imagery with machine-learning algorithms. The contribution of PIX4D and Picterra were prominent examples of contributions from the private sector to development aid.

More concretely, IOM and UNOSAT presented two projects applied to the Cox’s Bazar in Bangladesh to offer solutions to the Rohingya refugee crisis. They specifically showed how drone and satellite imagery combined with AI could help decision-makers to understand the size and complexity of a refugee camp half of the size of the city of Geneva. This is highly relevant and applicable to a myriad of other contexts. For instance, such technologies would allow for zooming in and out in conflict zones to design peace-keeping interventions; or in the case of natural disasters, prepare emergency responses and manage refugee flows.

While the Cox’s Bazar and similar examples are geographically specific challenges, EcoVision (ETH Zurich) demonstrated the vital role of satellite imagery in analysing the environment at a much larger scale. They showed that satellite data covering the entire globe could support scalable environmental programmes and push the envelope in terms of planning.

Finally, Tech4Impact (EPFL) presented its survey which identified 192 research labs that work on Sustainable Development, illustrating the existing supply of research and development that must match the demand from organisations.

Conclusion

An important lesson from this panel was the consensus among speakers that cutting-edge technologies, including AI and automation, can help aid and development actors do much more, much faster and with fewer resources – but such technologies cannot replace humans. Instead, they augment human capacity to tackle pressing issues and should be developed with this objective in mind.

An ongoing dialogue between scientists, practitioners and decision-makers is thus necessary to ensure that technology developments benefit society. The GSPI is committed to connecting organizations that need such technologies and experts who can provide them.