Smart24Food Atlas: Anonymized Dataset for Food Retrieval and Visual Portion-Size Matching Under Controlled Direct-Observation Conditions

This dataset contains the anonymized minimal data underlying the findings reported in the manuscript “Development of an Open-Access Food Atlas with Nutrient Composition and Recipes in Argentina: the Smart24Food Atlas.” It includes 2,160 participant–task observations from 108 participants completing 20 prespecified direct-observation tasks, comprising seven packaged-product selection tasks and 13 visual portion-matching tasks. The latter contribute 1,404 paired observations comparing the photograph-associated selected weight with the served weight. The workbook also includes participant-level data, task definitions, data dictionaries, food-specific descriptive summaries, analysis specifications and reported results. The dataset supports the reported analyses of food retrieval, exact matching, participant- and food-level variability, signed weight error, and descriptive Bland–Altman agreement under controlled direct-observation conditions. The dataset is anonymized and contains no directly identifying participant information. The data characterize performance for the specific food-retrieval and fixed-image matching tasks evaluated and should not be interpreted as validation of general usability or recall-based dietary assessment. The Smart24Food Atlas is available at https://smart24food.uap.edu.ar/atlas/