Oisin Mac Aodha

When Dr Oisin Mac Aodha completed his PhD at University College London (UCL) in 2013, there were lots of exciting ideas for machine learning applications, but few that worked. A conundrum that interested him was how AI could perceive and sense the world - it seemed almost impossible that computer vision could identify even a handful of objects from images. Over the next few years during a postdoc at UCL with Professor Gabriel Brostow and Professor Kate Jones, the boom in AI methods meant thousands of objects could actually be recognised. For Dr Mac Aodha, visual recognition tasks not only provide a rich view of the world but are also an excellent measure of the performance and advancement of AI.  

An algorithm only "sees" pixels in a photograph or video, there is no innate perception of the 3D shape of the scene or the depth of objects from the camera. During his postdoc at UCL, Dr Mac Aodha worked on methods to extract 3D understanding from images. In July 2016, Pokemon Go was launched. This augmented reality game used a smartphone camera to overlay virtual characters onto the real world that the player can hunt. But, because depth was not perceived accurately, characters did not interact with objects realistically, reducing enjoyment of the game. Niantic, the company that developed Pokemon Go, consulted with the team at UCL, including Dr Mac Aodha, to develop the core algorithms that solved the problem. This resulted in the spinout Matrix Mill, which was acquired by Niantic, igniting Dr Mac Aodha's innovation trajectory.  

During his postdoc at Caltech in Professor Pietro Perona's lab, between 2016 and 2019, Dr Mac Aodha worked on a project called Visipedia, with the goal of developing machine learning algorithms that can perform challenging tasks such as identifying animal species in photographs. This furthered his interest in the applications of computer vision to ecology research, which Dr Mac Aodha continued at the University of Edinburgh when he joined in 2019. Because conservation often focuses on specific species, ecological surveys usually involve specialists, they are time consuming and expensive to perform. They also do not easily capture ecosystem-level changes. Seeing an unmet need, Dr Mac Aodha, along with others in the field, identified that computer vision techniques offered a scalable method to collect ecology data and document where species are across the globe.  

Dr Mac Aodha partnered with iNaturalist. Drawing on the millions of images uploaded to this platform by citizen scientists, he developed algorithms to automatically identify species and predict their ranges (where a species is found). A notable benefit of Dr Mac Aodha's approach is to model species together, rather than creating algorithms for distinct species, meaning that changes over time can be tracked and eventually patterns and ecosystem-level connections can be uncovered. This allows investigation of the effects of climate change and other anthropogenic activities on biodiversity, as well as monitoring the impacts of conservation measures. Indeed, Dr Mac Aodha was awarded a Climate Change AI (CCAI) Innovation Grant in 2022 to support this work.  

Dr Mac Aodha explains: "The thousands of wildlife photos uploaded to the internet each day provide scientists with valuable insights into where different species can be found on Earth. However, knowing what species is in a photo is just the tip of the iceberg. These images are a hugely rich resource that remains largely untapped. Being able to quickly and accurately comb through the wealth of information they contain could offer vital clues about how species are responding to multi-faceted challenges like climate change." 

As an Aspiring Innovator, Dr Mac Aodha is highly collaborative, applying his expertise in many domains. For example, his lab also works on robotics, developing technology that allows machines to perceive their environment and safely interact with objects. He also firmly believes in AI with humans-in-the-loop: using machine learning as a tool to answer questions and learn. As such, he has partnered with the Cornell Lab of Ornithology to develop algorithms to help understand how a human learner can improve their knowledge of birds. A guiding theme of his research is to make AI techniques better - more robust, more reliable, and more trusted - so that it can be used to improve the environment and society. 

Dr Mac Aodha's advice to a researcher considering partnerships is: 

Look for those who complement what you are doing:  will integrating your research into their organisation result in a big impact? Can they illuminate new problems or research challenges that you were not previously aware of? Are they leaders in their respective areas? If the answer is yes, they may supercharge your work.