MétaCan
Menu
Back to cohort
Record W3118891363 · doi:10.2196/24467

The Human Factor in Automated Image-Based Nutrition Apps: Analysis of Common Mistakes Using the goFOOD Lite App

2021· article· en· W3118891363 on OpenAlexvenueno aff
Maria F. Vasiloglou, Klazine van der Horst, Thomai Stathopoulou, Michael P Jaeggi, Giulia S Tedde, Ya Lu, Stavroula Mougiakakou

Bibliographic record

VenueJMIR mhealth and uhealth · 2021
Typearticle
Languageen
FieldMedicine
TopicNutritional Studies and Diet
Canadian institutionsnot available
Fundersnot available
KeywordsComputer scienceArtificial intelligenceQuality (philosophy)Multimedia

Abstract

fetched live from OpenAlex

BACKGROUND: Technological advancements have enabled nutrient estimation by smartphone apps such as goFOOD. This is an artificial intelligence-based smartphone system, which uses food images or video captured by the user as input and then translates these into estimates of nutrient content. The quality of the data is highly dependent on the images the user records. This can lead to a major loss of data and impaired quality. Instead of removing these data from the study, in-depth analysis is needed to explore common mistakes and to use them for further improvement of automated apps for nutrition assessment. OBJECTIVE: The aim of this study is to analyze common mistakes made by participants using the goFOOD Lite app, a version of goFOOD, which was designed for food-logging, but without providing results to the users, to improve both the instructions provided and the automated functionalities of the app. METHODS: The 48 study participants were given face-to-face instructions for goFOOD Lite and were asked to record 2 pictures (1 recording) before and 2 pictures (1 recording) after the daily consumption of each food or beverage, using a reference card as a fiducial marker. All pictures that were discarded for processing due to mistakes were analyzed to record the main mistakes made by users. RESULTS: Of the 468 recordings of nonpackaged food items captured by the app, 60 (12.8%) had to be discarded due to errors in the capturing procedure. The principal problems were as follows: wrong fiducial marker or improper marker use (19 recordings), plate issues such as a noncompatible or nonvisible plate (8 recordings), a combination of various issues (17 recordings), and other reasons such as obstacles (hand) in front of the camera or matching recording pairs (16 recordings). CONCLUSIONS: No other study has focused on the principal problems in the use of automatic apps for assessing nutritional intake. This study shows that it is important to provide study participants with detailed instructions if high-quality data are to be obtained. Future developments could focus on making it easier to recognize food on various plates from its color or shape and on exploring alternatives to using fiducial markers. It is also essential for future studies to understand the training needed by the participants as well as to enhance the app's user-friendliness and to develop automatic image checks based on participant feedback.

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 imitation

Not 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.

metaresearch head score (Codex)0.006
metaresearch head score (Gemma)0.059
Version: metacan-v3-hybrid-931329e0061cValidation status: machine_predicted_unvalidated
Candidate categoriesnone
Consensus categoriesnone
DomainCandidate signal: none · Consensus signal: none
Study designCandidate signal: Observational · Consensus signal: Observational
GenreCandidate signal: Empirical · Consensus signal: Empirical
Teacher disagreement score0.006
Threshold uncertainty score0.031

Distilled classifier scores by category (both heads)

CategoryCodexGemma
Metaresearch0.0060.059
Meta-epidemiology (narrow)0.0010.000
Meta-epidemiology (broad)0.0010.001
Bibliometrics0.0020.001
Science and technology studies0.0010.001
Scholarly communication0.0020.002
Open science0.0010.002
Research integrity0.0010.001
Insufficient payload (model declined to judge)0.0010.001

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.

Opus teacher head0.070
GPT teacher head0.409
Teacher spread0.339 · how far apart the two teachers sit on this one work
Validation statusscore_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it

Classification

machine, unvalidated

Machine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.

The models applied no category: nothing in the taxonomy fit this work.
Study designObservational
Domainnot available
GenreEmpirical

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".

Quick stats

Citations23
Published2021
Admission routes1
Has abstractyes

Explore more

Same venueJMIR mhealth and uhealthSame topicNutritional Studies and DietFrench-language works237,207