Jove
Visualize
Contact Us

Related Experiment Video

Updated: Jun 27, 2026

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

Prompt Engineering and Model Selection for LLM-Based Nutritional Estimation from Food Images: A Multi-Dataset

Shinichi Nakagawa1, Akira Yamamoto2

  • 1Research Institute of Info-Communication Medicine (RinCOM), Tokyo 184-0004, Japan.

Nutrients
|June 26, 2026
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

Mechanistic Models: Compartment Models in Algorithms for Numerical Problem Solving01:29

Mechanistic Models: Compartment Models in Algorithms for Numerical Problem Solving

Mechanistic models play a crucial role in algorithms for numerical problem-solving, particularly in nonlinear mixed effects modeling (NMEM). These models aim to minimize specific objective functions by evaluating various parameter estimates, leading to the development of systematic algorithms. In some cases, linearization techniques approximate the model using linear equations.
In individual population analyses, different algorithms are employed, such as Cauchy's method, which uses a...

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Editorial for "In Vivo Microstructure Imaging in Oropharyngeal Squamous Cell Carcinoma Using the Random Walk With Barriers Model".

Journal of magnetic resonance imaging : JMRI·2023
Same author

The Efficacy of CT Temporal Subtraction Images for Fibrodysplasia Ossificans Progressiva.

Tomography (Ann Arbor, Mich.)·2023
Same author

Time-dependent diffusion MRI to distinguish malignant from benign head and neck tumors.

Journal of magnetic resonance imaging : JMRI·2018
Same author

Clinical evaluation of time-of-flight MR angiography with sparse undersampling and iterative reconstruction for cerebral aneurysms.

NMR in biomedicine·2017
Same author

Three-dimensional models of the segmented human fetal brain generated by magnetic resonance imaging.

Congenital anomalies·2017
Same author

Relationship between aging and T<sub>1</sub> relaxation time in deep gray matter: A voxel-based analysis.

Journal of magnetic resonance imaging : JMRI·2017
Same journal

RETRACTED: Dal Monte et al. Fatty Acids Dietary Supplements Exert Anti-Inflammatory Action and Limit Ganglion Cell Degeneration in the Retina of the EAE Mouse Model of Multiple Sclerosis. <i>Nutrients</i> 2018, <i>10</i>, 325.

Nutrients·2026
Same journal

RETRACTED: Cammalleri et al. A Dietary Combination of Forskolin with Homotaurine, Spearmint and B Vitamins Protects Injured Retinal Ganglion Cells in a Rodent Model of Hypertensive Glaucoma. <i>Nutrients</i> 2020, <i>12</i>, 1189.

Nutrients·2026
Same journal

RETRACTED: Locri et al. Protective Efficacy of a Dietary Supplement Based on Forskolin, Homotaurine, Spearmint Extract, and Group B Vitamins in a Mouse Model of Optic Nerve Injury. <i>Nutrients</i> 2019, <i>11</i>, 2931.

Nutrients·2026
Same journal

Correction: Pramesthi et al. Evaluating the Impact of Indonesia's National School Feeding Program (ProGAS) on Children's Nutrition and Learning Environment: A Mixed-Methods Approach. <i>Nutrients</i> 2025, <i>17</i>, 3575.

Nutrients·2026
Same journal

Correction: Huang et al. Correlation Study Between Dietary Behaviors, Lifestyle, and Psychological Problems in Chinese Children Aged 3-7. <i>Nutrients</i> 2025, <i>17</i>, 176.

Nutrients·2026
Same journal

Association of Dietary Animal and Plant Protein Composition with All-Cause Mortality: 24-Year Population-Based Cohort Study.

Nutrients·2026
See all related articles
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Large language models (LLMs) can estimate nutritional content from food images, with prompt design and model choice significantly impacting accuracy. Claude Sonnet provides the best value for nutritional estimation tasks.

Area of Science:

  • Artificial Intelligence
  • Computer Vision
  • Nutrition Science

Background:

  • Accurate nutritional content estimation from food images is vital for dietary assessment and public health.
  • Large language models (LLMs) show potential but require optimized prompt design and model selection for accuracy.
  • The impact of prompt engineering and model choice on LLM-based nutritional estimation is not well understood.

Purpose of the Study:

  • To evaluate the performance of three Claude models (Haiku, Sonnet, Opus) for visual nutritional estimation.
  • To compare the effectiveness of a default prompt versus a specialized visual estimation prompt.
  • To assess nutritional estimation accuracy across diverse food image datasets (meals and packaged foods).

Main Methods:

  • Three Claude models (Haiku, Sonnet, Opus) were tested on three datasets: NutriImage, SNAPMe, and JBFD.
Keywords:
ClaudeJBFDNutriImageSNAPMedietary assessmentfood imagelarge language modelnutritional estimationprompt engineering

Related Experiment Videos

Last Updated: Jun 27, 2026

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

  • A systematic comparison was made between a default prompt and a visual estimation prompt.
  • Five nutritional components (energy, protein, fat, carbohydrate, salt equivalent) were estimated.

    • Main Results:

    • The visual estimation prompt significantly improved accuracy with capable models (e.g., energy R² increased from 0.23 to 0.60 with Sonnet on JBFD).
    • Sonnet and Opus models outperformed Haiku, with minimal differences between Sonnet and Opus.
    • Packaged food images (JBFD) generally yielded better results than meal images; salt estimation remained challenging.
    • On the SNAPMe dataset, Sonnet demonstrated superior performance in estimating energy, protein, and fat compared to a previous ChatGPT-5 study.

    Conclusions:

    • Claude Sonnet presents an optimal cost-performance balance for LLM-based nutritional estimation.
    • Prompt design is crucial for accuracy, but its effectiveness is contingent on the model's inherent visual recognition capabilities.
    • The study underscores the complexity of visual nutritional estimation and offers practical insights for developing dietary assessment systems.