Capturing children food exposure using wearable cameras and deep learning
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Children's dietary habits are influenced by complex factors within their home, school and neighborhood environments. Identifying such influencers and assessing their effects is traditionally based on self-reported data which can be prone to recall bias. We developed a culturally acceptable machine-learning-based data-collection system to objectively capture school-children's exposure to food (including food items, food advertisements, and food outlets) in two urban Arab centers: Greater Beirut, in Lebanon, and Greater Tunis, in Tunisia. Our machine-learning-based system consists of 1) a wearable camera that captures continuous footage of children's environment during a typical school day, 2) a machine learning model that automatically identifies images related to food from the collected data and discards any other footage, 3) a second machine learning model that classifies food-related images into images that contain actual food items, images that contain food advertisements, and images that contain food outlets, and 4) a third machine learning model that classifies images that contain food items into two classes, corresponding to whether the food items are being consumed by the child wearing the camera or whether they are consumed by others. This manuscript reports on a user-centered design study to assess the acceptability of using wearable cameras to capture food exposure among school children in Greater Beirut and Greater Tunis. We then describe how we trained our first machine learning model to detect food exposure images using data collected from the Web and utilizing the latest trends in deep learning for computer vision. Next, we describe how we trained our other machine learning models to classify food-related images into their respective categories using a combination of public data and data acquired via crowdsourcing. Finally, we describe how the different components of our system were packed together and deployed in a real-world case study and we report on its performance.
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.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
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 it