General purpose AI models give mixed accuracy estimating eaten meal calories
Small Average Error, Wide Individual Error: General-Purpose LLMs Estimating Energy Intake from Meal Photos in a Feeding Trial
Human-Computer InteractionComputer Vision and Pattern Recognition
Summary
Estimating how much energy people get from meals just by looking at photos could help with diet tracking, but it’s tricky. The authors tested six big AI models on meal photos from a study where exact intake was measured. They found that these models often make big mistakes for individual meals, though their overall average error is small. Using a photo before and after eating didn’t hugely improve accuracy. The study shows these AI estimates shouldn’t replace dietitians but might help support their work.
What this means in practice
- •For dietitians and nutritionists: Support dietary assessment workflows by providing photo-based preliminary estimates of energy intake to complement professional judgment.
- •For health app developers: Incorporate AI-powered energy intake estimation from meal photos to enhance food tracking features with known limitations on individual accuracy.$Commercial implications: Enables development of photo-based dietary logging features that supplement manual calorie tracking in health and diet applications.
Authors
Xiaomeng Li
Abstract
Photo-based estimation of food energy could ease dietary assessment, and general-purpose multimodal large language models (LLMs) make such estimates easy to obtain. Most evaluations use one photo taken before eating and compare with energy served, while dietary assessment needs energy eaten. We evaluated six models from two vendors (Claude Haiku, Sonnet and Opus 5.5; GPT-6 luna, 6.1 sol and 6 astra) on 802 meals from ACETADA, a controlled feeding trial with weighed intake, with the before photo only, before and after photos in separate calls (subtraction), or both photos in one request. Three results ran against intuition. Subtraction under-estimated intake for five of six models, because the models still report hundreds of kilocalories for a nearly empty tray. Scaling the before-only estimate by the average share eaten (0.87) matched much of the gain from the after photo; beyond that constant, the second photo lowered meal-level error by only 2.2 to 3.4 points, mainly in meals with food left. With one photo, two models had a near-zero intake bias only because they under-estimated the served energy and ignored uneaten food. In ablations, adding the after photo improved accuracy under both prompt formats for five models, while prompt effects varied across models. Over 148 feeding days, three models kept the 90% confidence interval of their mean error within a pre-set 10% margin, but no model reached the pre-set 30% limits of agreement. Protein was under-estimated by every model, most in protein-dense meals; repeated calls changed error by at most 1.4 points; a high reasoning effort helped only the smallest model (4.2 points) and changed the others by less than 1 point. A low average bias is not evidence of accurate individual intake estimation; such estimates should support, not replace, assessment by a dietitian.