Skincare Recommendation System using

Computer Vision
GUIDE
Dr . J Cruz Antony, MCA.,Ph.D.

Team member 1 Team member 2

Name of the Student : Vikram Kumar.K Name of the Student : Emuri Bhavanesh
Register Number : 40731125 Register Number : 40731701

April 16, 2024 Department of CSE 1

 PRESENTATION OUTLINE
• Abstract
• Objective
• Introduction
• Problem Statement
• Proposed Solution
• Literature Survey
• Base Paper
• System Architecture
• Data Flow
• References
• Conclusion
April 16, 2024 Department of CSE 2
 ABSTRACT
In this presentation, we will explore how computer vision and artificial intelligence can be used to improve
skincare recommendations. We will discuss the importance of skincare and how it affects our overall health and
appearance. We will also define computer vision and explain how it can be used to analyse skin conditions and
make personalized skincare recommendations.
Additionally, we will delve into the role of artificial intelligence in skincare and highlight the benefits of using
technology to improve skincare, such as improved accuracy and personalized recommendations. By the end of
this presentation, you will have a better understanding of how technology can revolutionize the skincare
industry and help you achieve your best skin yet.

April 16, 2024 Department of CSE 3

 OBJECTIVE
The objective of this project is to introduce the audience to the exciting possibilities of using computer
vision and artificial intelligence in skincare. We will explore how these technologies can be used to
analyse skin conditions and make personalized skincare recommendations, leading to improved
accuracy and better results for users.
Through this presentation, we hope to educate the audience on the importance of skincare and how
it affects our overall health and appearance. By leveraging the power of technology, we can provide
customized solutions to address individual needs and preferences.

April 16, 2024 Department of CSE 4

PROBLEM STATEMENT
Skincare is a multi-billion dollar industry and yet, many people struggle to find products that work
for them. The problem lies in the fact that everyone's skin is unique and there are countless factors
that can affect its health and appearance.
Traditional skincare recommendations often rely on broad categories like 'oily' or 'dry' skin types,
which can be inaccurate and ineffective. This is where computer vision and artificial intelligence
come in - by analyz individual skin conditions and using machine learning algorithms, personalized
skincare recommendations can be made with much greater accuracy.

April 16, 2024 Department of CSE 5

 PROPOSED SOLUTION

• Our proposed solution involves using computer vision and artificial intelligence to analyse skin
conditions and make personalized skincare recommendations.
• By analyse factors such as skin type, age, and environmental exposure, our system can provide
tailored recommendations for each individual user. This approach not only improves the accuracy
of skincare recommendations but also saves time and money by avoiding trial-and-error with
different products.

April 16, 2024 Department of CSE 6

LITERATURE SURVEY
S.N TITLE AUTHOR YEAR Inference MERITS DEMERITS
O

1 SMART FACIAL SKINCARE Hung-Tse Chan, Ting-Yu Lin, 2021 The research paper highlights the he study presents a fresh business The research has a narrow scope, concentrating on
PRODUCTS USING Shih-Chun Deng, Chih-Hsien transformative impact of computer model using computer vision for skin type and acne, excluding other concerns.
COMPUTER VISION Hsia, and Chin-Feng Lai vision on personalized skincare skincare, enhancing consumer Comparative analysis is lacking, affecting the
TECHNOLOGIES recommendations. Machine learning relations. It delivers accurate skin evaluation of the system's efficacy and superiority
and deep learning enhance accuracy analysis, allowing personalized against existing recommendations.
over traditional methods. The system product recommendations, with
exhibits strong skin type classification superior type classification accuracy.
results, yet requires more research for
comprehensive comparison with
existing systems

2 Based on machine learning for Hsiao-hui Li , Yuan-Hsun 2020 Research proposes a promising ML- The study's holistic approach The research lacks extensive validation, requiring
personalized skin Liao , Yen-nun Huang , Po- based skincare recommendation combines image processing, object broader dataset testing for credibility. Additionally,
care products recommendation Jen Cheng system. YOLOv4 aids accurate detection, and multi-label the absence of comparisons with existing systems
engine personalized suggestions, yet classification for personalized hinders insights into its effectiveness.
validation and comparisons with others skincare recommendations.
are vital for establishing its value YOLOv4's accuracy identifies skin
conditions and prevents damage by
guiding to suitable products

3 Facial Skincare Products’ Ting-Yu Lin , Hung-Tse Chan 2022 Based on the experimental results, the Fast and Contactless Identification: Risk of Forgetting Passwords: One of the risks of
Recommendation with Computer , Chih-Hsien Hsia , and Chin- finger-vein identification system using The proposed finger-vein traditional identity identification methods is that
Vision Technologies Feng Lai the proposed SVM model and the identification system in this work is clients may forget their passwords
skincare products' recommendation designed to be fast and contactless,
system using ML and DL methods show allowing consumers to instantly
promising results in terms of accuracy verify their membership without the
and efficiency need for physical cards or
passwords

4 A Natural Language Processing Adelola Adebo 2020 The proposed model for personalized This enables the model to better One drawback of topic-based approaches,
Approach skincare recommendation systems capture word context information, including the proposed model, is their failure to
to a Skincare Recommendation shows promise in addressing the which has proven invaluable in capture word context information fully . While the
Engine limitations of existing approaches. By successful recommendation model incorporates review words, it may still not
incorporating different review elements, systems. capture the complete context of the words used in
deep neural networks, and considering the reviews
user opinions, the model aims to
improve the accuracy and effectiveness
of skincare recommendations.

7
April 16, 2024 Department of CSE
 BASE PAPER
Based on machine learning for personalized skin care products
recommendation engine

Authors : Hsiao-hui Li , Yuan-Hsun Liao , Yen-nun Huang , Po-Jen Cheng

Publication : 2020 International Symposium on Computer, Consumer and Control (IS3C)

Abstract :
This project focuses on leveraging machine learning and deep learning algorithms for an intelligent skin care
recommendation platform. In a diverse skin care market, personalized solutions are essential due to varying skin
conditions. The platform employs YOLOv4 for facial feature detection, analyzing sub-images for defects, and
correlating features. Skin condition classification is achieved through image preprocessing and a multi-label
model. The outcome is a predictive model that offers personalized skin care advice and product recommendations
based on individual skin conditions.
April 16, 2024 Department of CSE 8
SYSTEM ARCHITECTURE

April 16, 2024 Department of CSE 9

DATA FLOW

Computer Vision
start User Input Data Collection processing (Object
Detection)

Natural Language
Data
Processing
Preprocessing
(Text Data) Feature
Engineering

Deploy Feedback Machine

Personalization
(Output) loop Learning Model

April 16, 2024 Department of CSE 10

 TOPICS/MODULES
1.IMAGE ACQUISITION AND PREPROCESSING :
This module initiates the skincare analysis process by capturing high-quality images of the user's skin through the mobile app's
camera. These images undergo preprocessing, which includes standardizing lighting conditions and enhancing image quality
to ensure accurate analysis.

2.FACE DETECTION AND LOCALIZATION:

The "Face Detection and Localization" module precisely identifies and localizes the user's face within the captured image. This
critical step allows the system to concentrate its analysis solely on the facial region, ensuring the accuracy of subsequent
skincare assessments.

3.SKIN ANALYSIS:
At the heart of our system, the "Skin Analysis" module dissects the user's skin features. It segments the facial image into
distinct regions and employs advanced techniques, such as texture and colour analysis, to quantify various aspects, including
wrinkles, dark spots, pores, and redness, providing a comprehensive view of skin condition.

April 16, 2024 Department of CSE 11

 TOPICS/MODULES
4. SKIN TYPE CLASSIFICATION:
This module determines the user's skin type by leveraging the outcomes of the skin analysis. Based on the analysis results, it
classifies the user's skin into categories like oily, dry, combination, or sensitive, allowing for tailored product recommendations
aligned with the user's specific skin type.

5.INGREDIENT ANALYSIS:
"Ingredient Analysis" scrutinizes the composition of skincare products. It cross-references product ingredients with the user's
skin type and sensitivities, identifying allergens and irritants. Additionally, it highlights the benefits of specific ingredients in
addressing the user's unique skin concerns.

6. NLP FOR USER INPUT:

Through "NLP for User Input," the system processes and interprets user-generated text input. It employs NLP techniques to
understand skincare concerns, preferences, and queries, ensuring that the system responds effectively to the user's needs and
objectives.

April 16, 2024 Department of CSE 12

 TOPICS/MODULES
7. PRODUCT RECOMMENDATION ENGINE:
Our recommendation engine, the "Product Recommendation Engine," leverages the insights gathered from previous modules
to provide personalized skincare product suggestions. It factors in the user's skin type, specific skin conditions, ingredient
compatibility, and user input to deliver tailored recommendations, all while considering historical user data and product reviews.

8.FEEDBACK AND LEARNING:

The "Feedback and Learning" module continuously refines our skincare recommendation system. It actively collects and
analyses user feedback on recommended products, enabling the system to adapt and improve over time. Incorporating user
interactions and preferences, it enhances the accuracy and personalization of skincare advice. This iterative learning process
ensures that recommendations become increasingly effective and tailored to individual users.

April 16, 2024 Department of CSE 13

 WORKING PRINCIPLES
1.USER PROFILE AND INFORMATION : Users provide their personal information, including age, gender, and known skin

allergies, to create a profile within the system.

2. FACIAL IMAGE ANALYSIS: Computer vision techniques are employed to analyze uploaded or captured facial images,

extracting important features related to skin condition.

3. SKIN ANALYSIS: Machine learning models categorize the user's skin type and identify any skin abnormalities or specific

concerns, such as acne or wrinkles.

4. PRODUCT RECOMMENDATION: The system recommends skincare products or cosmetics that match the user's skin type

and address their specific needs, while also considering their allergy history.

5. USER FEEDBACK AND PRIVACY: Users can provide feedback on the recommendations, and strong privacy measures are

in place to protect sensitive user data throughout the process.

April 16, 2024 Department of CSE 14
 REFERENCES
1.Facial Skincare Products’ Recommendation with Computer Vision Technologies by Ting-Yu Lin
1,Hung-Tse Chan 2,Chih-Hsien Hsia 2,ORCID andChin-Feng Lai 1 : https://www.mdpi.com/1433664

2. SMART FACIAL SKINCARE PRODUCTS USINGCOMPUTER VISION TECHNOLOGIES Hung-Tse Chan1,

Ting-Yu Lin2, Shih-Chun Deng3, Chih-Hsien Hsia1,*, and Chin-Feng Lai21 Department of Computer Science
and Information EngineeringNational Ilan University, Taiwan2 Department of Engineering ScienceNational
Cheng Kung University, Taiwan :
https://ieeexplore.ieee.org/abstract/document/9689474

3. Based on machine learning for personalized skincare products recommendation engineHsiao-hui

LiBachelor's Degree Program in ChainStore ManagementTainan University of TechnologyTainan, TaiwanPo-Jen
ChengDepartment of Computer Science andInformation EngineeringNational Chung Cheng UniversityChiayi,
TaiwanYuan-Hsun LiaoīMaster Program of Digital InnovationTunghai UniversityTaichung, TaiwanYen-nun
HuangResearch Center for InformationTechnology InnovationAcademia SinicaTaipei, Taiwan:
https://ieeexplore.ieee.org/abstract/document/9394031/

April 16, 2024 Department of CSE 15

Conclusion

Skincare recommendations using computer vision offer personalized advice by analyze skincare
product images and user input. Computer vision techniques extract visual features, while machine
learning processes user data to generate tailored recommendations based on skin type, concerns,
and preferences. Continuous improvement through user feedback ensures up-to-date suggestions.
Challenges include data quality, privacy, and the need for professional dermatological advice.
Nevertheless, this innovative approach holds promise for promoting healthier skincare
experiences in the future.

April 16, 2024 Department of CSE 16