Automatic Image Recognition Meal Reporting Among Young Adults: Randomized Controlled Trial
Bibliographic record
Abstract
Background: Advances in artificial intelligence technology have raised new possibilities for the effective evaluation of daily dietary intake, but more empirical study is needed for the use of such technologies under realistic meal scenarios. This study developed an automated food recognition technology, which was then integrated into its previous design to improve usability for meal reporting. The newly developed app allowed for the automatic detection and recognition of multiple dishes within a single real-time food image as input. App performance was tested using young adults in authentic dining conditions. Objective: A 2-group comparative study was conducted to assess app performance using metrics including accuracy, efficiency, and user perception. The experimental group, named the automatic image-based reporting (AIR) group, was compared against a control group using the previous version, named the voice input reporting (VIR) group. Each application is primarily designed to facilitate a distinct method of food intake reporting. AIR users capture and upload images of their selected dishes, supplemented with voice commands where appropriate. VIR users supplement the uploaded image with verbal inputs for food names and attributes. Methods: The 2 mobile apps were subjected to a head-to-head parallel randomized evaluation. A cohort of 42 young adults aged 20-25 years (9 male and 33 female participants) was recruited from a university in Taiwan and randomly assigned to 2 groups, that is, AIR (n=22) and VIR (n=20). Both groups were assessed using the same menu of 17 dishes. Each meal was designed to represent a typical lunch or dinner setting, with 1 staple, 1 main course, and 3 side dishes. All participants used the app on the same type of smartphone, with the interfaces of both using uniform user interactions, icons, and layouts. Analysis of the gathered data focused on assessing reporting accuracy, time efficiency, and user perception. Results: For the AIR group, 86% (189/220) of dishes were correctly identified, whereas 68% (136/200) of dishes were accurately reported. The AIR group exhibited a significantly higher degree of identification accuracy compared to the VIR group (P<.001). The AIR group also required significantly less time to complete food reporting (P<.001). System usability scale scores showed both apps were perceived as having high usability and learnability (P=.20). Conclusions: The AIR group outperformed the VIR group concerning accuracy and time efficiency for overall dish reporting within the meal testing scenario. While further technological enhancement may be required, artificial intelligence vision technology integration into existing mobile apps holds promise. Our results provide evidence-based contributions to the integration of automatic image recognition technology into existing apps in terms of user interaction efficacy and overall ease of use. Further empirical work is required, including full-scale randomized controlled trials and assessments of user perception under various conditions.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.008 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.006 | 0.003 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.002 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.004 | 0.002 |
| Insufficient payload (model declined to judge) | 0.008 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".