Penn State researchers have developed an artificial intelligence system that counts how quickly children take bites during meals, achieving roughly 70% accuracy compared to human observers in a pilot study aimed at identifying obesity risk factors.
The system, called ByteTrack, addresses a key challenge in childhood obesity research: faster bite rates are associated with greater obesity risk, but measuring this behaviour requires researchers to manually watch and count each bite in video recordings, making large-scale studies impractical.
Yashaswini Bhat, doctoral candidate in nutritional sciences and lead author on the study, collaborated with Timothy Brick, associate professor of human development and family studies, to build the AI model. The system identifies children’s faces in videos featuring multiple people before detecting individual bites during eating.
The researchers trained ByteTrack using 1,440 minutes of video footage from 94 children aged seven to nine consuming four meals on separate occasions. After coding 242 videos manually to train the model, they tested it on 51 additional videos from the same dataset.
Results demonstrated the model was approximately 97% as successful as humans at identifying children’s faces in videos, but reached about 70% accuracy for bite identification. The system performed excellently when children’s faces remained in clear view, but struggled when faces were obscured or when children chewed on utensils or played with food.
Faster eaters at greater risk
Kathleen Keller, professor and Helen A. Guthrie Chair of nutritional sciences at Penn State and study co-author, explained the obesity connection. “When we eat quickly, we don’t give our digestive track time to sense the calories,” she said. “The faster you eat, the faster it goes through your stomach, and the body cannot release hormones in time to let you know you are full. Later, you may feel like you have overeaten, but when this behaviour repeats, faster eaters are at greater risk for developing obesity.”
Previous research from Keller’s laboratory demonstrated that faster bite rate, particularly combined with larger bite size, associates with higher obesity rates among children. Bite rate represents a stable characteristic of children’s eating behaviour that can be targeted to reduce eating speed and ultimately obesity risk.
Alaina Pearce, research data management librarian at Penn State and study co-author, noted that bite rate often becomes the target behaviour for interventions aimed at slowing eating rate.
Bhat outlined future applications for the technology. “The eventual goal is to develop a robust system that can function in the real world,” she said. “One day, we might be able to offer a smartphone app that warns children when they need to slow their eating so they can develop healthy habits that last a lifetime.”
The National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of General Medical Sciences, the Penn State Institute for Computational and Data Sciences, and the Penn State Clinical and Translational Science Institute funded the research, published in Frontiers in Nutrition.
