Glossary
Definitions in straightforward language for 15 terms related to nutrition science, metabolism, AI food tracking, GLP-1 medications, and dietary evaluation methods.
ai-tech
AI Food Recognition
AI food recognition employs computer vision and deep learning techniques to recognize foods from an image and assess their nutritional values. In calorie tracking applications, it allows users to take a picture of a meal and automatically receive dish identification, portion estimates, and calorie counts without the need for manual database searching.
Computer Vision
Computer vision refers to the domain of artificial intelligence that focuses on training software to analyze images and videos. In calorie tracking applications, computer vision facilitates AI food recognition, where a meal photograph is processed to predict the foods present and their quantities.
Food Classification
Food classification is an AI task that involves determining which dish or food appears in a photograph. In calorie tracking applications, food classification generates the dish label (for example, "grilled chicken breast," "caesar salad," or "pad thai") that is then matched to a database entry to obtain calorie and macro data.
Multimodal AI
Multimodal AI refers to artificial intelligence that processes various types of inputs, usually combining visual data (images) with linguistic data (text), and at times also incorporating audio or sensor information. In calorie tracking applications, multimodal AI represents the architectural evolution driving AI food recognition: the model can accept both a photograph and a textual description (such as "this is grilled chicken with rice") to yield a more precise dish identification and portion estimate than either input could achieve individually.
Portion Estimation
Portion estimation is the AI task of determining the quantity of food present on a plate from a photograph. In calorie tracking applications, portion estimation often accounts for the greatest source of calorie discrepancies, as visually similar plates can vary by 50% or more in actual weight due to dish density, concealed ingredients, and camera perspective.
database
Crowdsourced Database
A crowdsourced database is a food database where entries are primarily submitted, modified, or validated by app users instead of being managed by paid editorial staff or reference bodies like the USDA. These databases provide extensive coverage with millions of entries, albeit with variable quality and significant error rates in user-reported nutrition data.
USDA FoodData Central
USDA FoodData Central serves as the U.S. Department of Agriculture's official, publicly accessible database for nutritional composition of foods. It is the definitive reference for calorie and macronutrient content in whole foods within the United States and acts as the foundational database for Independent Reviews's accuracy testing procedures.
Verified Food Database
A verified food database refers to the curated selection of food entries within a calorie tracking app, where the nutritional information has been validated against a primary source, which is usually the manufacturer's product label, USDA FoodData Central, or the nutrition information provided by restaurants. Verified entries constitute the reliable core of any extensive food database, while unverified user-submitted entries comprise a much larger but less dependable margin.
accuracy
Dietary Assessment
Dietary assessment involves clinical and research methodologies that focus on quantifying what individuals consume. It includes techniques such as 24-hour recall, food-frequency questionnaires, weighed dietary records, and photo-based documentation. Calorie tracking applications act as consumer-oriented dietary assessment tools, and the academic literature on dietary assessment provides the methodological framework for their evaluation.
MAPE
Mean Absolute Percentage Error (MAPE) is the primary metric for assessing the accuracy of calorie tracking applications. It indicates how much an app’s calorie estimate diverges from the actual calorie content of a meal, represented as a percentage. A lower MAPE signifies a more precise application.
Mean Absolute Percentage Error
Mean Absolute Percentage Error is the extended term for MAPE, a statistical measure that quantifies accuracy in forecasting or estimation as the average percentage deviation between predicted and actual figures. In the context of nutrition application testing, mean absolute percentage error is the standard metric for illustrating how far a calorie tracker's estimates are from lab-measured true values.
Weighed Reference Meals
Weighed reference meals are testing meals whose actual calorie and macronutrient values are established through precise weighing of each component against the USDA FoodData Central database, rather than relying on estimates. They serve as the laboratory standard against which the estimates from calorie tracking applications are evaluated in our accuracy testing.
pricing
Free Tier
A free tier refers to the no-charge level within a calorie tracking app's pricing model. In the current 2026 applications, the free tier usually provides limited daily log capabilities, lacks AI photo recognition, offers restricted database access, and includes ads in the user experience, while advanced features are reserved for paid subscriptions that range from $40 to $120 annually.
Freemium
Freemium is the pricing strategy where an app's essential features are available for free, while more advanced features require a paid subscription. Each major calorie tracking application in 2026, including MyFitnessPal, Cronometer, Lose It, Cal AI, and MacroFactor, employs some variant of the freemium model, with annual subscriptions typically priced between $40 and $120.