Artificial intelligence and computer vision for quantitative assessment of eating behavior and urban food landscapes

What we eat and where we can buy food are key to our health, but studying these things has always been slow and expensive. Researchers spend weeks watching videos of people eating to count bites or walking through neighborhoods to check what food is available. My PhD explored how artificial intelligence can take over this work.
On a city-wide scale, I used online restaurant menus from Boston, London, and Dubai to map which neighborhoods have access to healthy food, finding that wealthier areas have more nutritious options. On an individual scale, I built AI models that watch videos of people eating and automatically count their bites, chews, and sips with high accuracy. These tools were combined into an open-source platform that cuts analysis time from six weeks to six hours.
By making nutrition research faster and more scalable, this work helps scientists and policymakers better understand and ultimately improve how people eat.