Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR921

Prediction of Probable Allergens in Food Items

Using Convolutional Neural Networks
Harshavardan. R. 1; Kanish. S.2; Madhav Suta Adityan. G.3; Rathi Gopalakrishnan4
Department of Computer Science and Engineering
Sri Ramakrishna Engineering College Coimbatore, India

Abstract:- Food monitoring and nutritional analysis play

a crucial role in addressing allergen-related health issues,
and their importance continues to grow in our daily lives.
In this study, we utilized a convolutional neural network
(CNN) to recognize and analyze food images, assess the
nutritional content of dishes, and provide information on
potential allergens. Identifying food items from images
poses a significant challenge due to the wide variety of
foods available. To address this, we leveraged the
Logmeal API, which utilizes CNN to identify various
types of meals, their ingredients, and potential allergens.

Keywords:- Convolutional Neural Network (CNN), Food

Image Recognition, Convolution Layers, Nutrition,Logmeal
API,Food Allergies

I. INTRODUCTION

Today, individuals are increasingly conscious of their

dietary choices as a means to prevent future or existing health
issues. Precise evaluation of ingredients and calorie content
is crucial for ensuring proper nutrient intake and mitigating
the risk of allergies. The accurate categorization of food
items holds significant importance in this regard. In recent
years, extensive research has been dedicated to leveraging
computer vision and artificial intelligence technologies for
the analysis of food images and automated vision-based
system for assessing traditional food involves four main
steps: food identification, food type classification, weight
assessment, and obtaining health information. Advancements
in image processing, object recognition, machine learning, Fig 1: System Architecture
and deep learning, particularly convolutional neural networks
(CNN), have significantly enhanced the accuracy of food A. Related Work
recognition applications. Utilizing the Logmeal API, we In recent years, deep learning and convolutional neural
classify food using CNN and store the data in FirebaseDB in networks (CNN) have been instrumental in advancing food-
JSON format for this project. related methods. The abundance of food images accessible on
the Internet, including platforms like Google image search,
has facilitated the aggregation of extensive data for training
deep learning models. Overcoming challenges unique to the
culinary domain, such as variation and similarity between
categories, necessitates the use of sophisticated and
intelligent algorithms. This section provides a review of
literature on food analysis, significant works on multimodal
learning and food, and the application of these techniques
within restaurant contexts.

IJISRT24APR921 www.ijisrt.com 1142

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR921

II. LITERATURE SURVEY

In [1], the survey lays the groundwork by providing

essential information on food allergies. This may include
coverage of topics such as allergens, immune responses,
diagnostic methods, and management strategies. In [2] and
[3], the paper concentrates on developing a framework that
employs deep learning techniques to real-time estimate the
nutritional value of food. It is likely to delve into the
methodology, data sources, deep learning models used, and
the accuracy of nutritional value estimation. In [4], the
paper likely presents an innovative approach that combines
fuzzy clustering and a Whale-based neural network for food
recognition and calorie estimation in daily dietary
Fig 3: Food Recognition
assessment. In [5], the paper introduces a new dataset
specifically tailored for food recognition tasks. The papers are C. Allergen Detection Module
likely to detail experiments using various recognition This module uses data obtained from Firebase DB and
algorithms, as well as discussing the outcomes, challenges,
Logmeal API to identify allergens in a specified food item. It
and potential enhancements in food recognition technology. involves comparing the ingredients of the food item with the
In [7][12], the paper describes the creation of a real-time food allergen data stored in the Firebase database to detect any
recognition system meant for smartphones, possibly matches.
examining the system architecture, image processing
techniques, machine learning models used, and real-world
D. User Interface Module:
performance metrics. In [8], the paper concentrates on The user interface module manages the user interaction
enhancing food recognition through the utilization of image
aspect of the project. It encompasses functions for capturing
augmentation techniques in combination with convolutional
user input (food images) and presenting the allergen
neural networks (CNNs), potentially covering different detection results to the user. React is utilized for the module
augmentation methods, CNN architectures, training to enhance the interface.
procedures, and improvements in recognition accuracy.
 CNN Architecture
III. MODULES The convolutional neural network comprises the input
layer, hidden layer, and output layer. Within the CNN, the
A. Firebase Integration Module:
intermediary layer is referred to as the hidden layer. In
The detection system relies on Firebase Realtime CNNs, the input is in the form of a tensor: (number of inputs)
Database, which provides functions for accessing the entire
x (input height) x (input width) x (input channel). Following
website to retrieve allergen data. It also allows for the
traversal through the convolutional layer, the image is
updating of allergen data with new information as required. consolidated into a feature map known as the Activation Map,
which is represented as: (Number of entries) x (height of
feature map) x (width of feature map) x (feature map
channel). The spiral layer may include pooling layers as well
as general and global layers in convolutional networks. The
subsequent layer of a single neuron comprises a group of
neurons that aggregate the data outputs with the scale of the
layer group. Known as the fully connected layer, it links each
neuron in one layer to every neuron in the following layer,
resembling the traditional multilayer perceptron (MLP)
neural network.

Fig 2: Fire Base DB

B. Logmeal API Integration Module:

The dataset comprises a variety of images featuring
diverse food items from various cuisines, obtained from the
Logmeal API for food recognition. Integration with the
Logmeal API enables the retrieval of food ingredient details.
The system includes functions to make HTTP requests to the
Logmeal API, parse the API responses, and extract pertinent
food ingredient information. Fig 4: CNN Architecture

IJISRT24APR921 www.ijisrt.com 1143

Volume 9, Issue 4, April – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165 https://doi.org/10.38124/ijisrt/IJISRT24APR921

 Allergens Display: allergen database to encompass a wide variety of allergens to

After completing the Deep Learning phase, the project adapt to new findings or advancements in allergen research.
enters the second phase, which involves displaying potential Also, to incorporate the application with wearable devices
food allergies. This is achieved by utilizing FirebaseDB to like smartwatches and trackers.
update real-time collected data and data obtained from
various internet sources in the form of a JSON file. Using REFERENCES
JavaScript, the database is connected to the existing Logmeal
API data. In the backend, a list of potential allergens and their [1]. “Food Allergies:The Basics” Division of
corresponding ingredients and allergies is updated. This Immunopathology, Department of Pathophysiology
allows for the display and sorting of potential allergies related and Allergy Research, Center for Pathophysiology,
to specific food items based on their ingredients. The process Infectiology and Immunology, Medical University of
of using FirebaseDB is as follows: Vienna, Vienna, Austria,2017
[2]. A Framework to Estimate the Nutritional Value of
 Retrieve Data Food in Real Time Using Deep Learning Techniques.
 Initializing the Database IEEE Access, 7 (1).
 Accessing Firebase Database [3]. Effective Learning and Classification using Random
 Processing and Filtering Data Forest Algorithm 1Vrushali Y Kulkarni, 2Pradeep K
 Integration with Food Detection Sinha International Journal of Engineering and
 Displaying Possible Allergens Innovative Technology (IJEIT) Volume 3, Issue 11,
May 2014.
IV. RESULTS [4]. W R SAM EMMANUEL and S JASMINE MINIJA”
Fuzzy clustering and Whale-based neural network to
We initiated our study by employing the CNN food recognition and calorie estimation for daily
architecture to recognize food images and identify potential dietary assessment” Sådhanå 2018.
allergens present in the food ingredients with the assistance of [5]. Amatul Bushra Akhil, Farzana Akter Tania Khatun &
FirebaseDB and LogmealAPI. As previously stated, the Mohammad Shorif Uddin “Recognition and
dataset utilized in the project yields an impressive accuracy Classification of Fast Food Images” Global Journal of
rate of 93%, delivering near-perfect results. Computer Science and Technology 2018.
[6]. Gianluigi Ciocca, Paolo Napoletano, and Raimondo
Schettini” Food Recognition: A New Dataset,
Experiments, and Results” IEEE 2017.
[7]. Real-time Food Recognition System on a Smartphone
Yoshiyuki Kawano · Keiji Yanai,2018
[8]. Image Augmentation-Based Food Recognition with
Convolutional Neural Networks.Lili Pan, Jiaohua Qin,
Hao Chen, Xuyu Xiang, Cong Li, Ran Chen
Computers, Materials &Continua 59 (1), 2019
[9]. Machine learning based approach on food recognition
and nutrition estimation Zhidong Shen, Adnan
Shehzad, Si Chen, Hui Sun, Jin Liu Procedia
Computer Science 174, 448-453, 2020
[10]. Food recognition for dietary assessment using deep
convolutional neural networks. Stergios
Christodoulidis, Marios Anthimopoulos, Stavroula
Mougiakakou .New Trends in Image Analysis and
Processing--ICIAP 2015
Fig 5: Output Design [11]. NutriNet: a deep learning food and drink image
recognition system for dietary assessment Simon
V. CONCLUSION AND FUTURE Mezgec, Barbara KoroušićSeljak
ENHANCEMENTS [12]. Food recognition accuracy in a mobile phone based
dietary assessment system Md Hafizur Rahmana,
In this initiative, our aim was to detect food allergies by Mark R Pickering, Deborah Kerr, Carol J Boushey,
analyzing the ingredients using various Deep Learning Edward J Delp 2012 IEEE International Conference.
algorithms. Leveraging Logmeal API and FirebaseDB, we
have effectively launched a system focused on food
allergens, which can benefit users seeking information about
the food they consume. For future improvements, we can
expand the database to identify a wider range of possible
allergens. To consistently update and broaden the known

IJISRT24APR921 www.ijisrt.com 1144