International Journal of Advanced Research and Publications

ISSN: 2456-9992

SMART - Machine Learning Based Fitness Mobile

Application
Santhiramohan Madhushika1, Mohomed Zowrie Mohomed Akil2, Nifthas A.R.M3, Pirtheep GT4, Ravi
Supunya Swarnakantha N.H.P5, Suriyaa Kumari6
1
Sri Lanka Institute of Information Technology, Faculty of Computing
Mount Vernon M.D Patana, Sri Lanka, PH-+94 77 895 9014
[email protected]
2
Sri Lanka Institute of Information Technology, Faculty of Computing
72/4C, Library Road, Maharagama, Sri Lanka, PH- +94 75 329 3186
[email protected]
3
Sri Lanka Institute of Information Technology, Faculty of Computing
239/2, muhaideen mp maawarthay block j east 2, Sammanthurai Sri Lanka, PH- +94 77 694 6266
[email protected]
4
Sri Lanka Institute of Information Technology, Faculty of Computing
New Kandy Road, Malabe, Sri Lanka, PH- +94 77 492 6126
[email protected]
5
Department of Information Technology, Sri Lanka Institute of Information
Technology, Matara, Sri Lanka
[email protected]
6
Department of Information Technology, Sri Lanka Institute of Information
Technology, Malabe, Sri Lanka
[email protected]

Abstract: Many people don’t pay attention to their health and fitness with their busy life styles. This situation vastly increased in Sri Lanka
with the economy crisis of the country. People spending a stressful life bad economy and shortage of products and services. Most people
spend time in the Fuel queues, and due to the shortage of fuel, they have trouble visiting fitness centers, etc. Meantime they miss fitness
guidance, motivations, and diet plans and face physical and mental troubles because of the lack of time. SMART mobile application is very
helpful for fitness users who really need to improve their health. SMART includes various aspects like workout, food schedule, and gym
equipment suggestion, and trainer help are covered. In today's world, mobile applications have become an indispensable tool for many
people. This is especially true in countries where other forms of technology may not be as readily available. For fitness enthusiasts looking
to save time and maintain their fitness goals, the SMART mobile app provides a secure and reliable solution. While the app is designed for
users with basic fitness knowledge, it also provides advanced features that cater to the needs of professional fitness coaches. The SMART
app leverages the power of machine learning, image processing, and the Python module to deliver a seamless user experience. With its
innovative technology and user-friendly interface, the SMART app is poised to revolutionize the way people approach fitness.

Keywords: Fitness, Mobile application, Workouts, Machine learning, Image processing.

1. Introduction wellness endeavors. A new fitness app has been designed for
Amidst the ongoing situation prevailing in the country, self-trained individuals who have experience in exercising
individuals are making a concerted effort to achieve their [1]. The proposed mobile application utilizes image
goals. Simultaneously, with the growing emphasis on recognition technology to accurately identify the user's
"maintaining good health" and "regular exercise" among fitness level, and provides personalized recommendations on
people, the fitness culture is gaining momentum, and an food and workout plans through the use of machine learning
increasing number of individuals are focusing on their algorithms and a recommendation system [1].
physique and even monitoring numerical aspects such as
body fat percentage [1]. Fitness apps are designed to enable The SMART app offers a comprehensive solution for fitness
individuals to compare their fitness performance with others, enthusiasts who are looking for timely workout routines,
providing motivation to improve their behavior [4]. This nutritional guidance, and access to virtual trainers. With a
feature allows users to view statistical graphs of physical wide range of gym equipment and workout plans, this mobile
activity among their friends or communities who also use the app is a must-have for anyone looking to maintain a healthy
app, encouraging them to strive for better results [4]. and active lifestyle.

By benchmarking their progress against others, users can The proposed system is designed to be user-friendly and
build stronger habits and enjoy the benefits of a healthier highly accessible. Upon accessing the system, the user will
lifestyle. Furthermore, these applications serve as a medium be prompted to answer a set of questions, and they will be
for people to interact with like-minded individuals who have able to upload or take an image using their own device. The
common fitness objectives, fostering a feeling of system will then analyze the image using image processing
camaraderie and encouragement throughout their physical techniques and provide the user with valuable information

Volume 6 Issue 5, May 2023 78

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

regarding workouts, gym equipment, and meal planning. The In 2020, a study was conducted to address security concerns
system aims to provide comprehensive guidance to the user, related to the protection of personal data in mobile fitness
including instructions on how to use the equipment correctly, applications. The researchers analyzed 110 healthrelated
how to plan meals effectively, and how to perform workouts applications and found that only 51 of them were deemed
correctly. Additionally, the system provides a list of the top trustworthy. To gather evidence, they examined the Wi-Fi
exercise machines and gym equipment recommended by communication utilized by these applications [6].
fitness coaches to help users select the best equipment for
their needs. In 2010, an application was developed that focuses on the
physical and mental health of undergraduate students at the
Overall, the system is an excellent tool for anyone looking to university. This application underwent comprehensive
maintain a healthy lifestyle and improve their fitness levels. analysis in this area [7]. An Android application was
The framework's user-friendly interface and comprehensive developed in 2018 to track fitness and provide advice based
guidance make it an indispensable tool for beginners and on fitness criteria. This application allows individuals to
experienced fitness enthusiasts alike. By leveraging the latest monitor their fitness progress [8].
advancements in image processing and fitness coaching, the
system provides valuable insights and suggestions to help The increasing reliance on mobile devices and apps in our
users achieve their fitness goals. daily lives has the potential to revolutionize the way we
approach our health. However, existing solutions that
The process for using the fitness-enhancing application is leverage sensor technologies to encourage healthier habits
relatively straightforward. The user needs to upload an image often struggle to motivate user participation. To address this
of the body part that they wish to improve. The system will challenge, we propose a novel approach in which fitness
then compare the image values with the available datasets challenges are integrated into users' daily routines to increase
using conventional neural network (CNN) to determine the engagement. To bring this concept to life, we introduce
appropriate type of meal that the user needs to consume GeoFit, a mobile app that allows users to discover and add
based on the provided instructions. Afterward, the system fitness challenges within their vicinity. GeoFit has the
will calculate the Body Mass Index (BMI) value for that potential to inspire and encourage users to make their health
specific body part image. a priority by offering a fun and competitive way to stay
active. This innovative fitness solution has the ability to help
This study aims to implement a supporting application that individuals seamlessly incorporate exercise into their daily
can significantly improve fitness by utilizing various routines [4]
machine learning algorithms and classification techniques to
provide the most accurate and effective solutions for users. 3. Methodology
By analyzing the user's body part images and providing
customized recommendations, the application can help users
achieve their fitness goals more efficiently. The use of
machine learning algorithms and classification techniques
allows the system to make personalized recommendations
for each user.

This enables the system to provide accurate and efficient

results, which can greatly enhance the user's fitness journey.
Moreover, the system can adapt to the user's changing needs
and preferences, ensuring that the recommendations remain
relevant and effective over time. In conclusion, the fitness-
enhancing application provides a user-friendly solution for
individuals seeking to improve their fitness. By using Figure 1: Overall System Diagram
machine learning algorithms and classification techniques,
the system can provide customized recommendations to The Proposed System, a supporting application for
users, which are tailored to their specific needs and enhancing fitness workouts, is depicted in Figure 1. The
preferences. The fitness industry could be transformed with mobile application aims to provide assistance and support for
this application, as it offers a highly effective and efficient fitness enthusiasts in improving their physical health. The
means for people to attain their fitness objectives. primary objectives of the application include helping users
with their workout routines, providing a product analyzer for
2. Background health supplements, and offering guidance on healthy fitness
If you're new to fitness, you can begin your journey by practices. The app is designed to be user-friendly and
enrolling in a group fitness class or hiring a personal trainer. accessible, catering to fitness users who are struggling to
As they become more acquainted with the equipment and navigate the complexities of maintaining a healthy lifestyle.
their own body, they may choose to work out on their own as
the cost of a personal trainer can be prohibitive [1]. The The proposed framework comprises of four main
responsibility of trainers includes keeping track of exercises, components that will be completed to achieve the desired
such as what movements to perform and how much weight to outcome. A range of CNN algorithms and models will be
use. When working out solo at a gym, many individuals employed. Once the model is trained, the binary classifier
choose to keep track of their workouts using a notebook or results will be sorted, and the frontend part implementation
mobile application [1]. will be carried out using Visual Studio. Python has been

Volume 6 Issue 5, May 2023 79

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

selected as the programming language, and several libraries individual system diagram for the recommendation of gym
will be used, including Image Processing and Machine equipment.
Learning frameworks. Additionally, a K-means clustering
model will be implemented in the development of the mobile 1.2 Getting information from users, and according to the
application. Image processing will be utilized to analyze the information system suggest meals for users
body part, which is an essential step in the disease detection
and severity assessment process. 1.2.1 Getting information from users
1) Recommendation of Gym equipment. This system is more dependable, secure, and user-friendly. In
2) Getting information from users, and according to the this component, the system will ask some important and
information suggest some meals for users. basic questions from users. What is your age? What is your
3) Personalized workout plan recommendation. height? What is your weight? Based on those values, the
4) Provide solutions, recommendations and ideas for system should analyze those things with existing data sets.
fitness and health related issues. Provide solutions using Following that, the system will allow fitness users to scan
machine learning and upload their body image into the system using their own
device’s camera to analyze the body part and give the value
Different types of CNN algorithms and models will be used of that particular body part based on an image processing
to examine the data analyzing. After the machine learning component. The system will compare the entered values to
model has been prepared, the binary classifier results are will the uploaded image based on those values.
be sorted. Visual studio used for frontend part
implementation. Python was picked as the programming The program will allow users to provide appropriate food
language., alongside the libraries, Machine Learning, Image items according to existing information, recommending food
Processing, Machine Learning framework, K-means for precise result. The primary goals of this application
clustering model to implementation of the mobile system are to provide food items with accurate information
application. Image processing for used to analyze the body and quality at the appropriate time, when user really need it,
part. to keep them from taking the incorrect food items to keep
their bodies and body part healthier. First, we should import
the data set, If the user needs to lose weight, the system uses
datasets to calculate their BMI values. The system will then
train those values before creating some array layers. This
array layer will be used to train the model using the
calculated BMI values. The system then recommends the
right and precise food to users in hopes of enhancing their
body part.

The system diagram in Figure 3 presents a typical

representation of the research component that includes
machine learning. The system works by first scanning the
image of a body part and processing it using image
processing techniques to separate the object. The K-means
Figure 2: Gym Equipment suggestion clustering algorithm is then used to identify the body part
type and analyze the image details with existing data and
1.2 Recommendation of Gym Equipment user input. This system can be particularly useful for
Systems recommends user requirements based on various individuals looking to improve their body fitness level by
criteria like given user information, user preferences, search utilizing a mobile application. With the ability to accurately
history, demographic information, user purchase history etc. identify body parts and analyze details, the system can
These kinds of recommendations are more important and provide valuable insights and recommendations to help users
beneficial when a person has no idea about a particular achieve their fitness goals.
selection / decision etc. A recommendation system is mostly
an Artificial intelligence or algorithm, usually combined with 1.2.2 K-Mean clustering algorithms.
machine learning. The K-means algorithm in data mining involves a series of
steps to analyze the learning data. Initially, a set of randomly
These recommendation systems are built to recognize user chosen centroids is selected as the starting points for each
characteristics, user preferences, user past decisions, by the cluster. Then, the algorithm performs iterative calculations to
interaction data gathered. They are impressions, likes, clicks, improve the position of the centroids by minimizing the sum
searches, and purchases etc. Recommendation systems are of squared distances between data points Figure 2: Gym
capable of providing highly personal predictions. equipment suggestion Figure 2: Gym equipment suggestion
Figure 3: Food suggestion part and their respective centroids.
Individuals that enjoy working out will use a variety of gym Through this process, K-means clustering helps to identify
equipment on their own. not just at home, but also in gym distinct groups or clusters within the data, which can aid in
facilities. It's advisable to be knowledgeable about the making predictions or uncovering patterns. By continuously
advantages of different gym equipment. Sometimes people refining the centroid positions, the algorithm seeks to find
become too attached to only one or two pieces of equipment the optimal clustering solution for the given data set.
and stop there. They can significantly improve if they have
the right equipment expertise. Below figure shows the

Volume 6 Issue 5, May 2023 80

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

Figure 3: Food Suggestion Figure 5: Personalized workout plan recommendation

The second step involves developing and training a machine

learning model using various algorithms such as regression
or neural networks. The model is trained using a combination
of supervised and unsupervised learning techniques to
predict the best workout routines for individuals based on
their BMI, age, and gender. The third step involves
evaluating the model's performance by comparing the
predicted workout routines with those that individuals are
currently doing or have done in the past. The model's
accuracy and effectiveness are assessed based on various
metrics such as precision, recall, and F1 score.
Figure 4: Sample K-Mean clustering Algorithm
The fourth step involves deploying the model to suggest
1.2.3 Data mining method workout routines to individuals based on their BMI, age, and
Data mining is a powerful tool that involves the application gender. The model can be deployed through a fitness app or
of statistical and computer methods to uncover valuable website, and individuals can input their personal details to
insights from large databases. By analyzing data from receive personalized workout recommendations. The fifth
various perspectives, patterns, correlations, and linkages that step involves monitoring and updating the model regularly to
were previously unnoticed can be discovered. By leveraging ensure that it remains accurate and effective. Any updates or
the power of data mining, companies can extract meaningful improvements to the model are made based on new data or
information from vast amounts of data, allowing them to feedback from users. The proposed methodology is expected
better understand customer behavior, identify market trends, to provide personalized workout recommendations to
and make data-driven decisions that lead to improved individuals based on their BMI, age, and gender. This
performance. As the volume of available data continues to approach is expected to lead to improved fitness outcomes
grow, data mining has become an indispensable tool for and reduce the risk of injuries or health issues. The model's
organizations across a wide range of industries. accuracy and effectiveness are evaluated using various
metrics, and any updates or improvements to the model are
1.3 Personalized workout plan recommendation made based on new data or user feedback.
The proposed methodology involves collecting and
preprocessing data, developing and training a machine And the proposed system takes in a video of an individual
learning model, evaluating the model's performance, performing a set of exercises and analyzes it using computer
deploying the model, and monitoring and updating the vision techniques. The system first identifies the individual's
model. The dataset used for training the model includes body parts and joints using a pose estimation model, such as
BMI, age, gender, and workout routines of individuals Open Pose. The system then calculates the angle and
collected from various sources such as health apps, and movement range of each joint during the exercise and
medical records. The proposed methodology is expected to compares it to ideal movement patterns for that exercise.
provide personalized workout recommendations to Based on this analysis, the system recommends a set of
individuals based on their BMI, age, and gender. exercises that are safe and effective for the individual's
specific needs. and Personalized workout recommendations
This methodology basically involves five steps: first one is based on BMI calculation, age, and gender using machine
data collection and preprocessing, model development and learning algorithms are expected to lead to improved fitness
training, model evaluation, model deployment, and model outcomes and reduce the risk of injuries or health issues.
monitoring and updating. The first step involves collecting
data from various sources such as fitness trackers and health 1.3.1 Deep Convolution Neural Network Design:
apps. The dataset includes BMI, age, gender, and workout The proposed deep convolutional neural network (DCNN)
routines of individuals. The data is then preprocessed to design takes in input images or videos of individuals
remove missing or incomplete data, check for outliers, and performing exercises and outputs a set of recommended
normalize the data. workouts based on the individual's BMI, gender, and age.
The network architecture consists of multiple convolutional

Volume 6 Issue 5, May 2023 81

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

layers. To prevent overfitting, the network includes dropout market. Dataset will consist of Age, gender, both mental and
layers that randomly drop out a fraction of the neurons physical health related issues, sleeping patterns, life style
during training. The output of the dropout layer is then patterns, workstyle and etc. Once the collection of data is
passed through a fully connected layer, which combines all done then data will be preprocessing the get specific and
the extracted features and outputs a set of recommended accurate results so the data will be cleaned which involves
workouts based on the individual's BMI, gender, and age. removing duplicates, correcting errors and filling missing
data points. Then in the second step, the model will be
1.4 Provide solutions, recommendations and ideas for developed and trained in order to produce correct results.
fitness and health related issues. Provide solutions using
machine learning.
In recent times there are several developments in the fitness
and health fields using machine learning process. There are
many notable success developments int the various topics
such as Customized workout plans can be made using
machine learning algorithms that consider a person's physical
attributes, fitness objectives, and past workout behavior.
Real-time feedback: Machine learning algorithms can
evaluate data from fitness trackers and other wearables to
give feedback on an individual's form, performance, and
efficiency while exercising.

1.4.1 Injury prevention

By analyzing a person's movement patterns and identifying Figure 6: System Diagram
potential injury risks, machine learning algorithms enable
early intervention and injury prevention. Nutritional advice: During this process the model will learn to make predictions
Based on a person's fitness objectives and an analysis of their by analyzing patterns in input and output data. Then the third
eating patterns, machine learning algorithms can provide step is evaluation of the model. In the fourth step the model
individualized recommendations for healthy eating. will be deployed. The deployment will be done in the form
of mobile application so it will be easy for users to access it.
1.4.2 Virtual coaching In the mobile application, the user can interact with the
Overall, implementing machine learning into fitness model to give inputs like age, gender, weight, height, health
alternatives has the possibility of significantly enhancing the related issues, lifestyle, workstyle & etc. Once the model
efficacy and customization of training programs, resulting in gets essential data, it will process the data and provide
better results for people. But most of the proposed systems information and solution like what might be possible health
and systems already available in the market for public usage related issues the user might face in the future, how to reduce
are mostly designed to work in a common way. Like only the chances of getting affected by those issues, how to
few common factors are considered and people do not have reduces the effects and symptoms of already existing issues,
the option to get specifically personalized services. what are the changes they have to make in their lifestyle,
Sometimes they might not get what they are looking for. what are the lifestyle changes they have to follow. In the fifth
Most of the available systems are giving limited amount of step, performance of the model will be monitored
information. They might not get solutions which including continuously and based on feedbacks from users, mandatory
all the aspects of health and fitness. updates and correction will be made to model and incase on
Figure 6: Individual System Diagram any updates the new
Overall, implementing machine learning into fitness data set also will be fed to the model to keep it updated.
alternatives has the possibility of significantly enhancing the
efficacy and customization of training programs, resulting in
better results for people. But most of the proposed systems
4. Results and Discussion
The mobile applications described in the above sections were
and systems already available in the market for public usage
developed using Android mobile application software. If you
are mostly designed to work in a common way. Like only
would like a modified version of the sentence, please let me
few common factors are considered and people do not have
know. One of the primary objectives of the solution that has
the option to get specifically personalized services.
been suggested is to enhance the level of fitness and
Sometimes they might not get what they are looking for.
workouts, gym equipment, food regimentation and mental
Most of the available systems are giving limited amount of
health, provide solutions, recommendations and ideas for
information. They might not get solutions which including
fitness and health related issues for fitness users. In
all the aspects of health and fitness.
conclusion, our study makes an important contribution to the
area of fitness by using this mobile application. The promises
This proposed system consists of five main steps are
of machine learning and image processing to improve fitness
including collection and preprocessing data, develop and
related issues and give proper suggestions to users. We think
train the model, verification and evaluation of the model and
our suggested method may be expanded upon to solve other
consistently monitoring and updating the model. In the first
issues in the area of online fitness and has important
step data will be collected from various sources to ensure
implementations for its long -term usability.
accuracy of the recommendations, data will be collected
from activity tracking apps, fitness apps and wide range of
both mental and physical health related apps available on the

Volume 6 Issue 5, May 2023 82

www.ijarp.org
 International Journal of Advanced Research and Publications
ISSN: 2456-9992

6. References
[1]. H. -Y. Kao and Y. -J. Lee, "Design and Implement a
Mobile Fitness Application based on Realtime Image
Detection," 2021 IEEE International Conference on
Consumer Electronics-Taiwan (ICCE-TW, Penghu,
Taiwan, 2021.

[2]. Jia-He Ye, Development of App for Strength Fitness

Products. 2017, Master Thesis, National Kaohsiung
University of Applied Sciences,
https://reurl.cc/KxXY2g .

[3]. D. Das, S. M. Busetty, V. Bharti and P. K. Hegde,

"Strength Training: A Fitness Application for Indoor
Based Exercise Recognition and Comfort Analysis,"
2017 16th IEEE International Conference on Machine
Figure 7: Home Page Figure 8: Workout Learning and Applications (ICMLA), 2017.
recommendation
[4]. A. Soontornwat, S. Funilkul and U. Supasitthimethee,
"Essential social attributes and Habit in fitness mobile
applications usage to motivate a physical activity,"
2016 International Computer Science and Engineering
Conference (ICSEC), Chiang Mai, Thailand, 2016.

[5]. V. Singh, A. Patade, G. Pawar and D. Hadsul,

"trAIner - An AI Fitness Coach Solution," 2022 IEEE
7th International conference for Convergence in
Technology (I2CT), 2022.

[6]. M. Parmar, S. Khant and A. Patel, "COVIFIT Mobile

Application for Fitness Improvement during Third
Wave of Covid19," 2022 2nd International
Conference on Innovative Practices in Technology
and Management (ICIPTM), Gautam Buddha Nagar,
Figure 9: Food Suggestion Figure 10: Image India, 2022.
uploading part
[7]. H. Haitao and D. Xinyan, "Development and
5. Conclusion Application of Undergraduate Physical Fitness and
The development of a mobile application for fitness Mental Health Detection System," 2010 Second
International Conference on Computer Modeling and
enthusiasts is a significant breakthrough in the world of
Simulation, 2010.
health and fitness. This app utilizes the mobile camera,
image processing, and machine learning modules to
[8]. R. Haji, S. Naik and R. Singh, "Fitness Tracking and
accurately identify the user's fitness level and recommend
Advisory Application," 2018 2nd International
customized workout plans. The primary focus of this study is
Conference on Trends in Electronics and Informatics
to improve the accuracy of fitness systems by gathering
(ICOEI), 2018.
image data through image processing. Utilizing this approach
would guarantee that users achieve more efficient outcomes
[9]. X. Yu, J. Yang, L. Luo, W. Li, J. Brandt and D.
and advantages from their workout routines. Additionally,
Metaxas, "Customized expression recognition for
integrating machine learning technology can significantly
performance-driven cutout character animation," 2016
improve the accuracy of activity tracking, yield valuable
IEEE Winter Conference on Applications of
insights into user behaviors, and produce custom-tailored
Computer Vision (WACV), Lake Placid, NY, USA,
workout plans. As this technology continues to evolve, the
2016.
potential for fitness apps to assist individuals in leading
healthier lifestyles is truly exciting. Overall, the development
[10]. Jia-He Ye, Development of App for Strength Fitness
of a mobile application for fitness enthusiasts that utilizes
image processing and machine learning technologies has the Products. 2017, Master Thesis, National Kaohsiung
University of Applied Sciences.
potential to revolutionize the fitness industry. By providing
accurate data and personalized recommendations, this app
[11]. M. Mardi and M. R. Keyvanpour, "GBKM: A New
can help individuals achieve their fitness goals more
Genetic Based KMeans Clustering Algorithm," 2021
efficiently and effectively.
7th In

Volume 6 Issue 5, May 2023 83

www.ijarp.org