About the Initiative
The Dietary Assessment Initiative is an independent, non-profit research collaboration studying the validity, accuracy, and reporting of dietary assessment instruments. Our scope spans the established self-report instruments (food-frequency questionnaires, 24-hour recalls, weighed-food records) and the newer generation of image-based and AI-mediated tools that are entering both consumer and clinical use.
The Initiative was assembled in 2023 by a small group of researchers who had worked separately on validation problems in nutritional epidemiology, dietetics, and computer vision, and who concluded — independently — that the field lacked an entity dedicated specifically to producing, replicating, and standardising accuracy evidence for these tools.
What we do
- Validation studies. Pre-registered, weighed-food-referenced evaluations of commercial and academic dietary assessment instruments.
- Systematic reviews. Quantitative syntheses of the published validation evidence for image-based and AI-mediated dietary assessment.
- Methodology briefs. Concise, citable explanations of agreement statistics, sample-size design, weighed-food protocols, and reporting standards.
- Datasets. Open release (where consent permits) of the reference meal sets used in our validation work.
- Commentary. Shorter, opinion-style writing on developments in the field.
What we do not do
- We do not certify products. We publish accuracy evidence; readers and clinicians use it as an input to their own decisions.
- We do not enter commercial NDAs. Where we evaluate a commercial product, we evaluate its publicly available surface as a black box.
- We do not accept funding from the products we evaluate. See our funding statement and conflict-of-interest policy.
Editorial conventions
Across publications we prefer Bland-Altman analysis and equivalence testing over null-hypothesis significance testing for agreement claims; report limits of agreement and intraclass correlation as primary outcomes for accuracy work; pre-register protocols where the design supports it; and label every reported number explicitly as either vendor-reported or independently-replicated.
Open access
All publications, methodology briefs, and dataset descriptors on this site are released under the Creative Commons Attribution 4.0 International License (CC BY 4.0). Reuse is encouraged with attribution.