AWS Pioneers Project

HALO secured a package of support from AWS, which could help it use a mix of AI and machine learning tools to detect and identify war debris using drone images from Ukrainian minefields, while giving them access to the highly secure and reliable AWS global cloud infrastructure to process and store the images.

The traditional method of discovering where minefields might lurk starts with a non-technical survey, which involves small teams visiting local communities to understand where accidents have happened or areas people avoid, to begin pinpointing where landmines might be situated. “It’s very painstaking and quite slow,” says Matthew Abercrombie, Head of Research and Development at HALO who has been leading this AI pilot effort.

In the last few years, the teams have made increasing use of drones, which take thousands of pictures that can then be stitched together to provide a detailed map for analysts to search for indicators of mines. “The problem is that an analyst searching through thousands of images takes a very long time. It can take three to five days to look through the imagery for an average minefield,” said Abercrombie.

“Working with AWS and using AI will transform the way we work. Taking the process of analysis down from five days to a ‘matter of hours’ is a game changer to detecting mines at speed,” says Abercrombie. 

His advice to other NGOs is to have a clear purpose about what they want AI to do. “Rather than throw a load of data at it, the key is to really focus in on a part of the pipeline that is slowing you down, a very manual process that involves data for instance.”

The other key piece of advice he has is to involve humans at every step. “The analysts on the ground are saying – this is what we are looking for, this is how we are looking for it, and these are the formats that we would need the data to be in at the end. Their opinions are hugely important to the success of this. Because if we made the perfect AI model and it didn’t give us the kind of insights that they needed, then they are never going to use it. And that is where the conversation with AWS started – they were saying, you need to tell us more about what the people on the ground are going to use this for.”

His third piece of advice is about getting the most out of the technology, within constrained budgets.

“The majority of our funding goes towards clearance, so we don’t have huge amounts of money to invest in new technologies. Partnering with AWS allows us to get the best from technology and really maximise the impact of the funding that we have.” That can mean helping educate marginalised communities who might not be plugged into the global conversation about technology.

“AWS provides the technology and then we can teach people how to use it to speed up demining and empower local communities.”

“Post-conflict communities are not places with huge amounts of job opportunities - so by taking local people, training them and teaching them how to do mine clearance work, we are giving them a stable income, which can be hugely transformative. If we use technology to accelerate the process, then that frees up time for people to be doing other jobs, whether that’s upskilling people on how to manage the machines and technology or teaching people IT skills.” He is incredibly proud of what the partnership has achieved so far, and what it can do in future.

“HALO has cleared over two million landmines globally and reclaimed over 760 square kilometres for safe use. As we expand into more advanced satellite imagery with AWS, we’re poised to uncover and clear even the most hidden hazards. Our ultimate goal is to speed up demining, empower communities, and ensure that each cleared square metre paves the way for community recovery and development.”