Virtual Apparel Try-on for web stores using Machine Learning

Project Title

Virtual Apparel Try-on for web stores using Machine Learning

Project Area of Specialization Augmented and Virtual RealityProject Summary

Ecommerce has evolved over the years and revolutionized retail. Online shopping is an ever-growing trend and

leading brands are striving to develop customer-friendly websites, offering a much richer buying experience. A

virtual try-on tool based on the concept of augmented reality is to be designed using machine learning algorithms

to provide the customer with an interactive experience of a real-world environment and adding digital imprints of

the products available to an image captured by a camera in real-time.

Project Objectives

Developing a customized Apparel web store integrated with a virtual try-on tool.

• Collecting sufficient data for the training and testing of the Machine Learning models.

• Training the model to identify the different nodes of a human body and various types of clothing items.

• Designing a computer vision-based interface to display the results via Augmented Reality

Project Implementation Method

The target subjects of our project are Web Development, Machine Learning, and Augmented Reality. Our website

is going to be designed in such a way that each user is going to be registered through a personal account, where

they can auto-save their shipping and billing addresses, payment methods, and clothing measurements.

Web Development:

In order to create an innovative and friendly user interface, we’re going to be using HTML, CSS, and JavaScript for

the front-end web development and as for the back-end web development, we plan on working with Python. We

have chosen Python as our primary language because Python coding is known to be very concise and

understandable and it has an abundance of libraries based on Machine Learning such as Numpy, Pandas,

TensorFlow, and Keras.

If a user wishes to try-on a piece of clothing before going through with an online purchase, our virtual try-on tool

will direct them to a page where they will be asked to provide access to their camera. For our project

demonstration, we’re going to use a 720p Logitech HD Camera as the image quality is an important factor for

image processing. On allowing access, a real-time image will be taken as input by the camera and the machine

learning algorithms will start to process the received data.

Machine Learning:

We have divided our machine learning tasks into three steps:

1. Collection of data

2. Data pre-processing

3. Model selection and testing

The data sets that are required for our project consist of images of body parts and various types of clothing items.

Data pre-processing is a technique used to analyze and label the raw data to make it suitable for training machine

learning models. This will help our model to identify the different nodes of a human body and correctly place the

piece of clothing where it belongs over the user input. The models that we plan on testing are CNN, YOLO v3,

YOLO v4, YOLO v5, and DNN. For model selection and evaluation, we will use Scikit-learn - a python-based

library which will assist us in feeding our clean data to each of the models and calculate their performance

accuracy.

Augmented Reality:

The output generated will be displayed on a screen at the user end via Augmented Reality by adding digital

content i.e., clothing imprints to real-life user input captured by a camera in real-time. We will deploy our website

using PaaS i.e., Platform as a Service, and our virtual try-on tool will be readily accessible to the users.

Benefits of the Project

An innovative e-commerce website designed with a virtual try-on tool that facilitates the users in

purchasing products online.

the end goal of our project here is to implement a virtual try-on tool which takes in-shop clothing and the customer’s image as input and then generates an output showing how it would look on them in real-time. This would allow users to shop with their family and friends from the comfort of their homes.

If this tool is implemented in a standard ecommerce website, it is expected to prompt a reduction in returns and exchange of products and increase sales and customer satisfaction. This would economically benefit the retailers and increase customer loyalty by many folds.

Technical Details of Final Deliverable

our proposed solution, the target subjects of our project are web development, machine learning and augmented reality.

Our website is going to be designed in such a way that each user is going to be registered through a personal account, where they can auto-save their shipping and billing addresses, payment methods, and clothing measurements. The coding for the front-end web development will be based on HTML, CSS, JavaScript. As for the back-end web development, we’re going to be using Python. What makes our website unique is our virtual try-on tool which we’re going to be integrating along with the well-known standard options such as ‘Add to cart’. Now as the users click on our virtual try-on tool, they will be directed to a page where they are asked to allow access to their camera. On allowing access, a real-time video will be taken as input by the camera and that’s when our machine learning algorithms will process the received data. We have divided our machine learning tasks into 3 steps. 1. Collection of Data set, 2. Data pre-processing and 3. Model selection/testing. Machine learning algorithms require a collection of data sets to help them identify and distinguish between the different nodes of a human body. This will help the model to display a digital form of clothing on top of the user input and make it seem as if he is actually wearing it in real-time. All of this will be made possible with the help of Augmented Reality.

Final Deliverable of the Project Software SystemCore Industry ITOther IndustriesCore Technology Augmented & Virtual RealityOther TechnologiesSustainable Development Goals Decent Work and Economic Growth, Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	6000
Logitech 720p HD Camera	Equipment	1	6000	6000