Analysing Skin and Recommending Makeup Products

Project Title

Analysing Skin and Recommending Makeup Products

Project Area of Specialization Artificial IntelligenceProject Summary

Nowadays everyone is familiar with the Make-up , it is used by many people as a beauty aid to improve or enchance their perceived attractiveness, cancel age cues and give them different looks for different ocassions. Make-up also bulids up self esteem and confidence. An estimated 44% of women do not leave their homes without make-up on. Those 44% of women believe that if they show their natural, untouched face, they won't be able to accomplish either of those things, and they will be treated differently. Research shows that foundation is the cosmetic that has the biggest impact on how women are perceived. This means that one should wear a perfect foundation shade of their skin tone in order to acheive a perfect make-up look. According to an informal survey in 2018, Toronto-based Make-up for Melanin Girls Founder Tomi Gbeleyi polled 5,500 women about the beauty industry. Gbeleyi found 80% of women faced challenges in finding a foundtion that matched their skin tone. According to a research group women spend $9.5 billion annually on beauty products and still many of them fail in getting make-up that matches their skin tone.

The project is set out to address problems women face in buying make-up products which match their skin tone. Mobile application is created which detects user's skin tone and skin texture and recommends the users a selection of make-up products according to their skin.

Project Objectives Project Objectives

1. The first objective that we want to accomplish in our project is to collect the dataset comprising of every skin tone and skin texture.

2. The second objective of our project is to label the dataset.

3. The third objective that we want to achieve is to train various algorithms.

4. The fourth Objective is to validate the results.
5. The fifth objective of our project is to create a mobile application that will allow the users to take a facial picture or import a facial picture for the detection of their skin tone and skin texture and recommendations of the cosmetic products.

Project Implementation Method

The project began with the research on human's different skin tones and skin textures. The dataset is collected of almost four thousand images comprising of different skin tones and skin textures. The dataset is then labelled using a Machine Learning algorithm that is "Self-Supervised Learning". Self-Supervised Learning also known as Self-Supervison, is a Machine Learning approach where the model trains itself by leveraging one part of data to predict the other part and generate labels accurately. As for our project, the one prior part of data is labelled by Sana'z Beauty Salon (located in Hyderabad, Sindh) to predict the remaining data in order to label the dataset autonomously.

Self-Supervised Learning reduced the time and cost to build the ML model. After getting the labelled dataset, the Face Detection Algorithm is applied on the dataset in order to avoid the background and to get the maximum accuracy. The skin tone and skin texture classification algorithm is applied in which color model RGB is converted into HSV color model and a standard thershold value is set to all input images. By using Deep Learning Algorithms ( Convolutional Neural Network), skin tone and skin texture is detected with maximum accuracy. The model is then tested to check/analyse if the model is working properly and accurately. In the last step/stage, the cosmetic/make-up products will be recccomended to the users according to their skin tones and skin textures.

Benefits of the Project

The project will reduce the time people spend in buying the make-up products. This project will also reduce the problem of spending money on foundations in order to get the perfect foundation shade that matches their skin tone and that hides their skin textures. By using the mobile application, the users will easily shop make-up products online without worrying. The project will provide the recommendations of cosmetic products from local brands to high-end brands of almost every range. The project will be gender flexible, every gender can use it to get the results of their skin tone, skin texture, and recommendations of make-up products. The project will increase the cosmetic Industry, Infrastructure and Innovation.

Technical Details of Final Deliverable

The final Deliverable will be the mobile apllication in which the users can take the facial image by using their mobile's camera or import flacial image from their mobile's storage. The Face detection algorithm will be applied on that picture to avoid background. Convolutional Neural Network (CNN) will analyse the skin tone and skin texture. Then the user will obtain the results of their skin tone and texture detected from the picture inserted. A selection of make-up products will be provided from local brands to high-end brands that will match the user's skin tone.  

To run CNN deep learning algorithm at large scale dataset a GPU is required.  GPUs are better at handling multiple but simpler calculations in parallel. NVIDIA provides something called the Compute Unified Device Architecture (CUDA), which is extremely beneficial for the computation of deep learning models. Therefore, a GPU with 16 GB RAM is used to run the CNN model.

Final Deliverable of the Project Software SystemCore Industry OthersOther IndustriesCore Technology Artificial Intelligence(AI)Other TechnologiesSustainable Development Goals Gender Equality, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
PNY NVIDIA Quadro P620 V2 Graphics Card VCQP620V2-PB	Equipment	1	50000	50000
Corsair VENGEANCE RGB PRO 16GB (2 x 8GB) DDR4 DRAM 3600MHz C18 Memory	Equipment	1	20000	20000