A PRELIMINARY REPORT

ON

NutriCheck: Decoding everyday products for optimal living

SUBMITTED TO THE SAVITRIBAI PHULE PUNE UNIVERSITY, PUNE

IN THE PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE AWARD OF THE DEGREE

OF

BACHELOR OF ENGINEERING (ARTIFICIAL INTELLIGENCE AND DATA SCIENCE)

SUBMITTED BY

STUDENT NAME EXAM SEAT NO:

Aniket Prakash Bhand BEAD21155

Nilkanth Dadaji Ahire BEAD21217
Shreyash Dinraj Chavan BEAD21157
Vaibhav Abhiman Jadhav BEAD21158

DEPARTMENT OF ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

DR. D.Y.PATIL INSTITUTE OF ENGINEERING, MANAGEMENT & RESEARCH

AKURDI, PUNE 411044

SAVITRIBAI PHULE PUNE UNIVERSITY

2024 -2025
1
 CERTIFICATE

This is to certify that the project report entitles

“NutriCheck: Decoding everyday products for optimal living”

Submitted by

STUDENT NAME EXAM SEAT NO

Aniket Prakash Bhand BEAD21155

Nilkanth Dadaji Ahire BEAD21217
Shreyash Dinraj Chavan BEAD21157
Vaibhav Abhiman Jadhav BEAD21158

is a bonafide student of this institute and the work has been carried out by him/her under the
supervision of Mrs. Sunayana Sutar and it is approved for the partial fulfillment of the requirement
of Savitribai Phule Pune University, for the award of the degree of Bachelor of Engineering
(Artificial Intelligence and Data Science).

Mrs. Sunayana Sutar Dr. Suvarna Patil

Guide Head
Department of AI&DS Department of AI&DS

Dr. Anupama V. Patil

Principal,
Dr.D.Y.Patil Institute of Engineering,Management & Research Akurdi Pune – 41

Place : Pune
Date :
2
 ACKNOWLEDGEMENT

It gives us great pleasure in presenting the preliminary project report on ‘Focus Feed: A Personalized
News Network’.

I would like to take this opportunity to thank my internal guide Mrs. Sunayana Sutar for giving me
all the help and guidance I needed. I am really grateful to them for their kind support. Their valuable
suggestions were very helpful.

I am also grateful to Dr. Suvarna Patil, Head of Artificial Intelligence and Data Science Department,
Dr. D. Y. Patil Institute of Engineering, Management & Research for her indispensable support,
suggestions.

In the end our special thanks to Mrs. Sneha Kanawade for providing various resources such as
laboratory with all needed software platforms, continuous guidance, for Our Project.

Aniket Bhand
Nilkanth Ahire
Shreyash Chavan
Vaibhav jadhav

(BE Artificial Intelligence and Data Science)

3
 ABSTRACT

The understanding of product labels is a challenge that everyone faces today especially when health
is of concern. Edible aimed to address this issue by facilitating the parsing of ingredient lists and
nutritional information. NutriChek is a web-based application, which is capable of immediate
analysis and interpretation of nutritional information off the labels of products by using Optical
Character Recognition combined with cutting edge Machine Learning and Natural Language
Processing techniques. The system works in the following way: text is extracted from an image, then
analyzed and a short report is generated that lists the benefits of the product, warns about the harmful
components together with their risks and proposes healthier substitutes. This functionality motivates
users to effortlessly reach the various health-related targets they set for themselves. In addition to
that, NutriCheck helps to provide the additional benefit of ingredient analysis by encouraging the
everyday evaluation of products and enabling the consumers to choose healthier options. Whether
you are out shopping or examining products at home, NutriCheck offers an easy and quick way to
help you stick to your health aspirations with your purchases and lead a healthier life, helping you
manage your purchases in relation to your health aspirations.

Keywords: Optical Character Recognition (OCR), Machine Learning, Large Language Models
(LLMs), Healthy Choices, Ingredient Analysis, Daily Products, Healthier Living.

4
 TABLE OF CONTENTS

LIST OF ABBREVATIONS i
LIST OF FIGURES ii
LIST OF TABLES iii

Sr. No. Title of Chapter Page

No.
01 Introduction 10
1.1 Overview 10
1.2 Motivation 10
1.3 Problem Definition and Objectives 10
1.4 Project Scope 11
02 Literature Survey 12
03 Software Requirements Specification 14
3.1 Assumptions and Dependencies 14
3.2 Functional Requirements 17
3.2.1 System Feature 1(Functional Requirement) 17
3.2.2 System Feature2 (Functional Requirement) 17
3.3 External Interface Requirements (If Any) 18
3.3.1 User Interfaces 18
3.3.2 Hardware Interfaces 18
3.3.3 Software Interfaces 18
3.3.4 Communication Interfaces 18
3.4 Nonfunctional Requirements 18
3.4.1 Performance Requirements 18
3.4.2 Safety Requirements 18
3.4.3 Security Requirements 18
3.4.4 Software Quality Attributes 18
3.5 System Requirements 19
3.3.1 Database Requirements 19

5
 3.3.2 Software Requirements (Platform Choice) 19
5.3.3 Hardware Requirements 19
3.6 Analysis Models: SDLC Model to be applied 19
3.7 System Implementation Plan 21
04 System Design 22
4.1 System Architecture 22
4.2 Data Flow Diagrams 23
4.3 Entity Relationship Diagrams 25
4.4 UML Diagram 26
4.5 Sequence Diagrams 26
4.6 Activity Diagram 27
05 Other Specification 28
5.1 Advantages 28
5.2 Limitations 28
5.3 Applications 28

06 Conclusions & Future Work 29

Appendix A: 29

Problem statement feasibility assessment using, satisfiability analysis and NP Hard,NP-

Complete or P type using modern algebra and relevant mathematical models.

Appendix B:

Details of the papers/Copyright. Summary of the Paper/copyright in not more than 3-4
lines. Here you should write the seed idea of the papers/ copyright you had referred for
preparation of this project report in the following format.
Example:
Thomas Noltey, Hans Hanssony, Lucia Lo Belloz,”Communication Buses for Automotive
Applications” In Proceedings of the 3rd Information Survivability Workshop (ISW-2007),
Boston, Massachusetts, USA, October 2007. IEEE Computer Society.

Appendix C: Plagiarism Report

References 29

6
 LIST OF ABBREVATIONS

Abbreviation Illustration
NCF Neural Collaborative Filtering
CBF Content-Based Filtering

NLP Natural Language Processing

TF-IDF Term Frequency-Inverse Document Frequency

7
 LIST OF FIGURES

FIGURE ILLUSTRATION PAGE NO.

1 Agile Model 20

2 System Architecture 22

3 DFD- level 0 23

4 DFD- level 1 23

5 DFD- level 2 24

6 Entity Relationship Diagram 25

7 UML diagram 26

8 Sequence Diagram 26

9 Activity Diagram 27

8
 LIST OF TABLES

TABLE ILLUSTRATION PAGE NO.

4.1 Literature Survey 12

9
 01. INTRODUCTION
1.1 OVERVIEW
Knowing the contents of products is a thing of a healthy lifestyle, but translating complexity in
product labels into an easily readable message for consumers is not always feasible. "NutriCheck:
Decoding Everyday Products for Optimal Living" helps out with this by offering a web-based tool
that uses advanced Optical Character Recognition (OCR), Natural Language Processing (NLP),
and Machine Learning to automatically decode and analyze ingredient lists. NutriCheck offers a
natural artwork on the labels of products that is transformed into easy-to-understand reports. The user
can fast-read for health risks and spot useful ingredients while giving consumers the suggestion of
healthier alternatives. It empowers consumers to make informed choices very easily toward a
healthier lifestyle while at the same time promoting greater transparency in everyday products.

1.2 MOTIVATION
In the health-conscious world today, consumers look for more transparency in what they consume
daily. However, product labels are so tough to decipher; complicated ingredient lists filled with many
scientific jargons most people aren't familiar with make it very confusing and lead to poor decision-
making on what's considered better. The problem is exacerbated by the vast number of products on
the market, so it will be near impossible for consumers to research any product they may encounter.
This was the inspiration behind the creation of NutriCheck: the need for a streamlined system that
would take real-time ingredient analysis and make it actionable. Using advanced OCR, Machine
Learning, and NLP technologies, NutriCheck, therefore, allows users to easily read what is written
on a product label, in turn enabling making better choices for an ultimately better, healthier lifestyle.

1.3 PROBLEM STATEMENT AND OBJECTIVE

Problem Statement:
Such long and indistinguishable ingredient lists on mundane products can mislead consumers from
getting their hands on harmful substances or even understanding the health benefits. Most of such
confusion disables an individual's better judgment toward making better decisions. In the light of
high number of products, it has become impractical to research each ingredient one at a time, which
creates a present demand for a tool that can simplify ingredient analysis and demonstrate clear,
actionable insights to improve decision-making processes.
10
Objectives:

• Accurate Text Extraction: To utilize Optical Character Recognition (OCR) technology to

extract text accurately from product labels and ingredient lists for further analysis.
• Ingredient Analysis: To analyze extracted ingredients for potential health risks and benefits,
identifying harmful substances while highlighting positive attributes.
• User-Friendly Interface: To develop an intuitive and easy-to-use web interface that allows users
to upload images and receive clear reports on product ingredients quickly.
• Healthier Alternatives Recommendation: To provide users with personalized
recommendations for healthier product options based on the analyzed ingredient data.
• Consumer Education: To enhance consumer awareness and understanding of product labels,
empowering individuals to make informed and healthier lifestyle choices.
• Data Insights for Brands: To generate valuable insights from ingredient analysis that can be
shared with businesses, helping them improve product formulations and align with consumer
health trends.

1.4 PROJECT SCOPE

Project Scope:
NutriCheck is a focus, on the development of an interactive web-based tool with a capability to help
consumers make better healthy lifestyle choices by making it easier for them to read product
ingredient labels. Leverage state-of-the-art Optical Character Recognition (OCR), Machine Learning
(ML), and Natural Language Processing (NLP) technologies and NutriCheck will enable users to
upload images of product labels with ease while the OCR component accurately extracts all the text
from such images, permitting smooth analysis of ingredient lists.
After extracting the text, the system will analyze the contents item by item to find hazardous
ingredients and whether such products pose health benefits. Users will therefore receive a thorough
report detailing dangers related to particular ingredients, mentioning useful components, and giving
alternatives to healthier consumption. The above feature promotes proactive nutrition; it makes it
look as if consumers are going to take control of everyday products they consume with great
reasoning.

11
 02. LITERATURE SURVEY

Sr Paper Title Year Dataset Methodology Findings Limitations

no

1. Innovative AI 2024 1176 FOP Object 94.9% Data from high-

methods for images detection and accuracy in market-share
monitoring from OCR for identifying brands only.
front-of- Amazon infant foods certification
package marks.
information

2. Sensitivity 2022 Chemical Global Sobol' and Limited real-

Analysis as an and quantitative FAST methods world model
Ingredient of environme SA methods improve model application
Modeling ntal policy independentce. discussion
models

3. Suspect and 2024 31 UNIFI 99 substances Only three

Non-Target household software for identified in substances had
Screening of cleaning screenings, screenings risk assessments
Chemicals in products. ECHA
Household database
Cleaning
Products

4. Multimodal 2024 High- Combines Multimodal Focused on

Fine-Grained resolution image and approach specific grocery
Grocery images OCR text outperforms product
Product and OCR data unimodal categories
Recognition text models

5. Consumer- 2024 602 Modified High CBLE Limited to

Based Label consumer CBLE scale labels Scandinavian
Equity for responses for label influence consumers
EDC-Free from performance consumer
Products Scandinavi behavior more.
an
countries

12
6. Online 2024 SQL/NoSQL OCR, NLP, Automated High costs,
Analysis of ingredient and LLMs ingredient barcode
Ingredient database. for safety analysis dependency,
Safety analysis enhances underutilized
consumer LLMs
safety

7. Health Risks of 2019 Data from Categorizes Identifies Many chemicals

Chemicals in 342 articles chemicals by widespread are under-
Consumer on 202 use and chemicals researched
Products chemicals. toxicity. like
phthalates
and BPA.

8. Hazardous 2021 Multiple Reviews Fragrances Insufficient

Effects of studies on fragrance cause contact fragrance
Fragrances in fragrance risks, dermatitis ingredient
Consumer allergens. including despite labeling and
Products dermatitis. benefits. regulation

9. High 2021 SHEDS-HT SHEDS-HT 60% of Regulatory data

Throughput model and and USEtox combinations needs
Risk Screening CPD models for have hazard improvement,
of Chemicals database. exposure. quotients >1. neglects exposure
variation.

10. Single Model 2022 CHEMDNE SCIBERT for Near state-of- Slightly lower F1
for Organic and R corpus chemical the-art scores than state-
Inorganic with 84,355 NER tasks. performance of-the-art
Chemical NER annotations. in chemical
NER.

Table 2.1: Literature Review

13
 03. SOFTWARE REQUIREMENT SPECIFICATION

3.1 INTRODUCTION
3.1.1 PROJECT SCOPE
• Ingredient Analysis and Reporting: Analyze extracted ingredient data to identify harmful
substances and assess potential health benefits, generating comprehensive reports for users..
• Personalized Health Recommendations: Provide users with tailored suggestions for healthier
alternatives based on the analysis of ingredient lists.
• User-Friendly Interface: Design an intuitive web interface that facilitates easy navigation and
interaction for users of all backgrounds, ensuring accessibility.
• Educational Resources: Incorporate informative content that enhances user understanding of
ingredient labels and nutrition, promoting informed decision-making.
• Scalability: Develop the system to efficiently handle a high volume of users and data,
ensuring consistent performance during peak traffic periods.

3.1.2 USE CLASSES AND CHARACTERISTICS

Our system is divided into following class/modules:
1. End Users: Consumers who upload images of product ingredient lists for analysis.
2. Administrators: Individuals managing user accounts and overall system functionality.
3. Content Providers: Contributors offering educational materials about nutrition and ingredient
safety.

3.1.3 ASSUMPTIONS AND DEPENDENCIES

Assumptions:
• User Engagement: Users will actively upload images of ingredient lists and provide feedback,
allowing the system to refine its analyses and recommendations effectively.
• Internet Access: The system requires users to have a stable internet connection to process
images in real-time and access the analysis results.
• Content Availability: Continuous access to updated databases of ingredients and their health
impacts is crucial for providing accurate assessments and suggestions.
• Device Compatibility: The platform is designed to be accessible on various devices (e.g.,
mobile, desktop, tablet), assuming users have compatible technology for optimal
functionality.

14
 Dependencies:
• Reliable OCR Technology: NutriCheck depends on OCR technology to read text from
ingredient images accurately. If the OCR doesn’t perform well, it could lead to incorrect
analyses.
• Machine Learning Libraries: We use machine learning libraries like TensorFlow and PyTorch
to process the text and generate health assessments. Their performance is crucial for accurate
results..
• Database Infrastructure: A strong and scalable database is needed to store user profiles and
interaction history. This ensures users can easily access their saved analyses.
• User Interface Frameworks: NutriCheck uses web development frameworks such as React
and Django to create a friendly and easy-to-navigate interface. Smooth performance is
important for keeping users engaged.

3.1.4 MATHEMATICAL MODELING

The mathematical modeling for NutriCheck revolves around two primary components: Optical
Character Recognition (OCR) and Natural Language Processing (NLP) integrated with machine
learning techniques. This combination allows for efficient extraction and analysis of ingredient data
from user-uploaded images, ultimately providing personalized health insights.

1. Optical Character Recognition (OCR)

OCR is the first step in our process, converting images of ingredient lists into machine-readable text.
The effectiveness of OCR can be mathematically represented through the following concepts:

• Image Input Matrix (I): An image III is represented as a matrix where each pixel's value
indicates intensity or color. The OCR process transforms this matrix into a textual
representation TTT by identifying and extracting characters and words.

• Character Recognition Model: The OCR system utilizes a machine learning model (often a
Convolutional Neural Network, CNN) to predict the likelihood P(c∣I)P(c | I)P(c∣I) of a
character ccc given the image input III. The output is a sequence of recognized characters
forming the ingredient list.

15
2. Natural Language Processing (NLP)
After OCR successfully extracts text, NLP techniques analyze the ingredients to provide
meaningful health insights:

• Feature Extraction: Each ingredient is represented as a vector VVV in a high-dimensional

space using methods like TF-IDF (Term Frequency-Inverse Document Frequency). The
importance of each term within the ingredient list is calculated, leading to a feature vector
representation:
V=TF-IDF(I)

• Health Assessment Model: We utilize machine learning models to evaluate the safety and
healthiness of extracted ingredients. This can be expressed as:
H=f(V)H = f(V)H=f(V)

3. Hybrid Model for Recommendations

The final stage combines insights from OCR and NLP to provide users with personalized health
recommendations. The overall prediction for health insights can be expressed as:
Ru=αH+(1−α)P

Here, RuR_uRu is the final recommendation for user uuu, HHH is the health assessment derived
from the NLP model, PPP represents user preferences or historical data, and α\alphaα is a tuning
parameter that balances these contributions.

Explanation

• OCR efficiently transforms ingredient images into actionable text, ensuring accurate data
input.
• NLP analyzes the textual data to provide insights regarding health implications, using both
feature extraction and predictive modeling.
• By integrating these components, NutriCheck delivers accurate, personalized reports on the
healthiness of everyday products, empowering users to make informed dietary choices..

The hybrid approach improves recommendation quality, ensuring that FocusFeed delivers timely,
personalized, and diverse news content.

16
3.2 FUNCTIONAL REQUIRMENTS

1.User Profile Management

• The system should allow users to create, update, and manage their profiles, including their
dietary preferences and health goals.
• Users must have the option to save their ingredient analysis history for future reference and
easy access.

2. Image Upload and OCR Processing

• The system must enable users to upload images of product ingredient lists effortlessly.
• Upon upload, the system should process the image using Optical Character Recognition
(OCR) to extract text from the ingredient list accurately.

3. Health Insight Generation

• The system should analyze the extracted ingredients using Natural Language Processing
(NLP) to provide detailed health insights.
• The insights should include potential health benefits, concerns regarding specific
ingredients, and recommendations for healthier alternatives.

4. User Interaction History

• The system must maintain a history of user interactions, including uploaded images and
generated reports, accessible in a "Saved Posts" section.
• Users should be able to view, edit, and delete entries from their interaction history at any
time.

• 5. Personalized Recommendations

• Based on user profiles and past interactions, the system should offer personalized
recommendations for healthier products and alternatives.
• Recommendations must update dynamically as users upload new images and provide
feedback on the insights received.

17
 3.3 EXTERNAL INTERFACE REQUIREMENTS
3.3.1 USER INTERFACES
• A clean, user-friendly dashboard where users can upload product images, view ingredient
analyses, and access their saved history.
• An admin panel for system administrators to manage updates, monitor performance, and
review system health.

3.3.2 HARDWARE INTERFACES

• Must support desktops, tablets, and smartphones to cater to a wide range of users

3.3.3 SOFTWARE INTERFACES

• APIs for real-time data from news sources.
• Integration with databases for storing user profiles and interaction history securely..

3.3.4 COMMUNICATION INTERFACES

• Secure communication protocols (such as HTTPS) must be used to ensure data privacy
and integrity.

3.4 NON-FUNCTIONAL REQUIREMENTS

3.4.1 PERFORMANCE REQUIREMENTS
• Ingredient analysis and health insights should be generated within 3 seconds.
• The system must be able to handle high user traffic without significant slowdowns.

3.4.2 SAFETY REQUIREMENTS

• Regular data backups must be performed to prevent data loss in the case of system failure.
3.4.3 SECURITY REQUIREMENTS
• All user data must be encrypted during storage and transmission.

3.4.4 SOFTWARE QUALITY ASSURANCE

• Rigorous testing procedures should be followed to ensure system reliability, accuracy,
and security.

18
3.5 SYSTEM REQUIREMENTS

• Database: A scalable database for storing user preferences and content data.
• Software: Use of Python for algorithms, Django for backend, and HTML, JavaScript for
frontend development.
• Hardware: Cloud-based infrastructure to support real-time recommendations.

3.6 ANALYSIS MODELS: SDLC MODEL TO BE APPLIED

The Agile SDLC model will be adopted to allow for iterative development, incorporating
continuous user feedback and system improvements throughout the project lifecycle. This
model ensures flexibility and keeps the project aligned with user needs and evolving
requirements.

19
Figure 3.6.1: Agile Model
20
3.7 SYSTEM IMPLEMENTION PLAN
Phase 1: Requirement Gathering and Analysis
• Objective: Collect detailed system requirements from stakeholders.
• Key Tasks: Identify user needs, analyze similar systems, define success criteria.
• Outcome: Comprehensive requirement specification document.
Phase 2: System Design
• Objective: Create system architecture and design for the recommendation engine,
database, and user interface.
• Key Tasks: Design the hybrid recommendation engine, create database schemas, design
user interfaces and data flow diagrams.
• Outcome: Complete system architecture and design documentation.
Phase 3: Development
• Objective: Build core system components, including recommendation engine and
article summarization.
• Key Tasks: Implement the NCF and CBF models, build user interfaces, integrate news
APIs, develop admin panel.
• Outcome: A functional prototype of FocusFeed ready for testing.
Phase 4: Testing
• Objective: Ensure system functionality, performance, and security.
• Key Tasks: Conduct unit, integration, performance, usability, and security testing.
• Outcome: Fully tested system, free of critical bugs, ready for deployment.
Phase 5: Deployment
• Objective: Deploy the system into the production environment.
• Key Tasks: Deploy on cloud infrastructure, set up CI/CD pipelines, ensure cross-
platform compatibility.
• Outcome: Live version of FocusFeed, accessible to users and providers.
Phase 6: Maintenance and Updates
• Objective: Continuously monitor and improve the system.
• Key Tasks: Monitor performance, fix bugs, provide updates, optimize security.
• Outcome: A maintained and evolving system with regular updates and high user
engagement.

21
 04. SYSTEM DESIGN

4.1 SYSTEM ARCHITECTURE

Figure 4.1.1: System Architecture for FocusFeed

22
4.2 DATA FLOW DIAGRAMS

Figure 4.2.1: DFD Level 0

Figure 4.2.2: DFD Level 1

23
Figure 4.2.3: DFD Level 2

24
4.3 ENTITY RELATIONSHIP DIAGRAMS

Figure 4.3.1: Entity Relationship Diagram

25
4.4 UML DIAGRAM

Figure 4.4.1: UML Diagram

4.5 SEQUENCE DIAGRAM

Figure 4.5.1: Sequence Diagram

26
4.5 ACTIVITY DIAGRAM

Figure 4.6.1: Activity Diagram

27
 05. OTHER SPECIFICATION
5.1 ADVANTAGES
• Personalized Ingredient Analysis: Delivers tailored insights based on the user's health
preferences and uploaded product ingredients, enhancing decision-making for healthier
choices.
• Enhanced Product Understanding: Utilizes OCR and LLMs to provide a detailed
breakdown of product ingredients, allowing users to understand even complex labels quickly
and easily.
• Time-Saving Insights: Generates concise, easy-to-understand summaries of product
ingredients, ideal for users looking to make quick, informed decisions.
• Real-Time Updates: Processes images of ingredient lists in real-time, providing immediate
feedback on product suitability and potential healthier alternatives.
• Scalability: Built on cloud infrastructure, NutriCheck can efficiently scale to support a
growing user base and increasing product uploads without affecting performance.
• User Engagement and Retention: Offers personalized recommendations and saves user
interaction history, encouraging users to return and fostering long-term engagement.
• Health-Conscious Recommendations: Provides users with healthier product alternatives,
supporting their journey towards making more informed and nutritious choices.

5.2 LIMITATIONS
• OCR Accuracy: The quality of ingredient recognition may be impacted by poor image
resolution or unclear product labels, leading to less accurate analysis.
• Dependency on Language Models: The system's responses may vary in accuracy
depending on the complexity of the ingredient list and the performance of the underlying
language models.
• Limited Data Sources: The tool relies on predefined data sets for ingredient analysis and
health recommendations, which may not cover all products or regional variations.
• User Input Quality: The accuracy of the results is highly dependent on the quality of the
images uploaded by users, which may lead to incorrect recommendations.
• Processing Time: While designed for real-time processing, large images or highly detailed
ingredient lists may take slightly longer to analyze, affecting user experience.
28
 06. CONCLUSION & FUTURE WORK
Conclusion:
NutriCheck is an innovative web tool designed to simplify ingredient analysis and promote healthier
lifestyle choices. By utilizing OCR and advanced language models, the system decodes complex
ingredient lists and provides users with clear, actionable insights. The tool empowers users to make
informed decisions about the products they consume, offering healthier alternatives when necessary.
With features like ingredient history tracking, NutriCheck not only educates users but also helps them
develop long-term, health-conscious habits in a world of often overwhelming product choices.
Future scope:
• Expanded Language Support: Enhance the OCR and LLM modules to support multiple
languages for a global user base.
• Mobile App Development: Develop a dedicated mobile application for easier access and
ingredient scanning on the go.
• Health Recommendations: Integrate personalized health recommendations based on user
dietary preferences or restrictions.
• AI-Driven Insights: Utilize advanced AI models to provide more detailed insights into potential
health risks of specific ingredients.
• Integration with Wearables: Sync with health-monitoring wearables to offer real-time dietary
advice based on user health data.

References:
1. Kim D, Kim S-Y, Yoo R, Choo J, Yang H (2024) Innovative AI methods for monitoring front-of-package
information: A case study on infant foods. PLoS ONE 19(5): e0303083.
https://doi.org/10.1371/journal.pone.0303083
2. Saltelli, A., et al. “Sensitivity Analysis as an Ingredient of Modeling.” Statistical Science, vol. 15, no. 4,
2000, pp. 377–95. JSTOR. Accessed 17 Oct. 2024.
https://www.jstor.org/stable/2676831
3. Kim JH, Kim TS, Yoon HJ, et al. Health risk assessment of dermal and inhalation exposure to deodorants
in Korea. Sci. Total Environ. 2018;625:1369–1379.
https://doi.org/10.4491/eer.2023.123
4. Pettersson, Tobias, Maria Riveiro, and Tuwe Löfström. "Multimodal fine-grained grocery product recognition
using image and OCR text." Machine Vision and Applications 35.4 (2024): 79.
https://github.com/Tubbias/finegrainocr.

29
5. Perkovic, Sonja. "Utilizing consumer-based label equity to signal consumer products free from endocrine-
disrupting chemicals." Journal of Retailing and Consumer Services 76 (2024).
10.1016/j.jretconser.2023.103611
6. V. C P, A. D, D. D. Kedilaya, S. S. Gondkar and S. Halhalli, "Online Analysis of Ingredient Safety,
Leveraging OCR and Machine Learning for Enhanced Consumer Product Safety," 2024 2nd International
Conference on Artificial Intelligence and Machine Learning Applications Theme: Healthcare and Internet
of Things (AIMLA), Namakkal, India, 2024, pp. 1-6, doi: 10.1109/AIMLA59606.2024.10531558.
https://ieeexplore.ieee.org/document/10531558
7. Wu, F., Li, Dingsheng, and Sangwon Suh. "Health risks of chemicals in consumer products: A review."
Environment international 123 (2019): 580-587.
https://doi.org/10.1016/j.envint.2018.12.033
8. Pastor-Nieto, María-Antonia, and María-Elena Gatica-Ortega. "Ubiquity, hazardous effects, and risk
assessment of fragrances in consumer products." Current treatment options in allergy 8 (2021): 21-41.
https://link.springer.com/article/10.1007/s40521-020-00275-7
9. Kim D, Kim S-Y, Yoo R, Choo J, Yang H (2024) Innovative AI methods for monitoring front-of-package
information: A case study on infant foods. PLoS ONE 19(5): e0303083.
https://doi.org/10.1371/journal.pone.0303083
10. Wu, F., Li, Dingsheng, and Sangwon Suh. "Health risks of chemicals in consumer products: A review."
Environment international 123 (2019): 580-587.
https://doi.org/10.1016/j.envint.2018.12.033

30