kth.sePublications KTH
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Automated Dietary Analysis Using Computer Vision and Large Language Models: An iOS Prototype
KTH, School of Engineering Sciences in Chemistry, Biotechnology and Health (CBH), Biomedical Engineering and Health Systems, Health Informatics and Logistics.
KTH, School of Electrical Engineering and Computer Science (EECS).
2025 (English)Independent thesis Basic level (degree of Bachelor), 10 credits / 15 HE creditsStudent thesisAlternative title
Automatisk kostanalys med datorseende och stora språkmodeller : En iOS-prototyp (Swedish)
Abstract [en]

Maintaining awareness of dietary habits is important for health, but traditional methods such as manual food journaling are time-consuming and often inaccurate. Advances in computer vision and language models offer opportunities to automate this process, making dietary assessment more convenient and reliable.

This thesis presents a prototype iOS application in which an on-device computer vision model extracts ingredients from meal photos and creates structured meal entries without manual typing. To turn these entries into practical guidance, the system aggregates meals over selected time windows and uses a large language model to assess dietary patterns, highlight benefits and risks, and propose targeted recommendations.

An evaluation based on meal plans aligned with dietary guidelines indicates that the prototype can identify overconsumption of discretionary foods and partly detect the absence of seafood, demonstrating the potential of combining computer vision with language models to support automated dietary assessment.

Abstract [sv]

Att ha medvetenhet om sina kostvanor är viktigt för hälsan, men traditionella metoder såsom manuell kostregistrering är tidskrävande och ofta opålitliga. Framsteg inom datorseende och språkmodeller erbjuder möjligheter att automatisera denna process och därigenom göra kostbedömningar mer lättillgängliga och tillförlitliga.

Denna rapport presenterar en prototyp av en iOS-applikation där en datorseendemodell på enheten identifierar ingredienser från matbilder och skapar strukturerade måltidsregistreringar utan manuell inmatning. För att omvandla dessa registreringar till praktisk vägledning sammanställer systemet måltider över valda tidsintervall och använder en stor språkmodell för att analysera kostmönster, framhäva fördelar och risker samt föreslå riktade rekommendationer.

En utvärdering baserad på måltidsplaner i linje med kostriktlinjer visar att prototypen kan identifiera överkonsumtion av energirika livsmedel som sötsaker och snacks samt delvis upptäcka avsaknaden av fisk och skaldjur. Detta demonstrerar potentialen i att kombinera datorseende med språkmodeller för att stödja automatiserad kostbedömning.

Place, publisher, year, edition, pages
2025.
Series
TRITA-CBH-GRU ; 2025:308
Keywords [en]
Vision-Based Dietary Assessment (VBDA), Semantic Segmentation, Computer Vision, Deep Learning, LLM, Mobile Application
Keywords [sv]
Vision-Based Dietary Assessment (VBDA), Semantisk segmentering, Datorseende, Djupinlärning, LLM, Mobilapplikation
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:kth:diva-370493OAI: oai:DiVA.org:kth-370493DiVA, id: diva2:2001430
Educational program
Bachelor of Science in Engineering - Computer Engineering; Master of Science in Engineering - Information and Communication Technology
Supervisors
Examiners
Available from: 2025-09-30 Created: 2025-09-26 Last updated: 2025-09-30Bibliographically approved

Open Access in DiVA

Automated Dietary Analysis Using Computer Vision and Large Language Models(2264 kB)125 downloads
File information
File name FULLTEXT01.pdfFile size 2264 kBChecksum SHA-512
c50aca2515bbe40729236af0838b2945a96d51b66c37dc43371cb8f42ee7ccc15140cb44f5eecaf3e0e579d4bb841deb544352e59b9b1b4ab25981f63253d9fc
Type fulltextMimetype application/pdf

By organisation
Health Informatics and LogisticsSchool of Electrical Engineering and Computer Science (EECS)
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 385 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf