According to a report by Statista Research Department, the global fashion eCommerce market stood at 700 billion dollars in 2021, and it is expected to reach 1.2 trillion by 2025. Although the market size is tremendously growing in the backdrop of huge advancements in technology and a post-pandemic world, online shoppers are faced with a new challenge of choosing the correct sizes and right fits when it comes to buying ready-made clothing.

This blog discusses a POC that has been designed by our Data Science experts to answer this dilemma for people who like to buy articles of clothing online. Dive in to learn more about our AI-powered virtual body measurement solution.

One major challenge for users who purchase apparel online is choosing the correct size from the available options for themselves. This is so because local and international brands tend to define sizes differently. The sizes vary by clothing item, brand, country, and also by time. In short, there is no standard way of measuring body sizes as there are no universal body types, and therefore differences arise.

Not to mention, having to measure yourself manually with a tape and keep track of your weight fluctuations only increases the number of steps one needs to take to make a purchase. This can easily cause frustration and compel a person to give up altogether on buying the item online.

You may also be interested in this case study “An Online Hotel Booking Platform Boosts Conversions with Our Python-based, ML Solution.”

Numerous methods have been suggested over the years to predict body measurements automatically by taking 3D pictures while using specialized equipment such as in-depth cameras. Although 3D-based technologies offer considerable accuracy, not all customers who shop online can use them owing to the availability factor. On the other hand, not much research has been conducted on the accuracy of 2D photos in estimating body measurements.

Royal Cyber data experts have endeavored to develop a smartphone system with the help of Artificial Intelligence, which will enable users to assess their body measurements and predict their body sizes by taking a 2D image from a standard smartphone camera.

The proposed approach uses body parts segmentation through computer vision and then applies pose detection to refine the measurements of specific body parts. For the system to do its job, you have to take your picture from specific dimensions, for instance, front view, side view, etc. Our solution can accurately measure body areas like the chest, waist, and inseam on the basis of captured images.

We have utilized computer vision and Machine Learning to estimate human body measurements and predict overall body sizes from 2D images taken from regular smartphones. In the following section, we will discuss the measurement technique in detail by going over the entire process in a step-by-step manner:

Firstly, in order to find body measurements, we need to detect the user’s silhouette or body shape by separating the user’s body from the background in the image.

Once the user’s body silhouette is obtained, we segment the silhouette into separate body parts so that we can identify where each part is located.

Along with body parts segmentation, we need to know what are the landmarks of human body parts. For example, we determine where the left and right shoulder points are and where the right or left waist points lie. This is done to further refine our measurements later.

We can filter the areas of interest to process only those parts which are important in making measurements. For instance, we can keep our focus on the waist or abdomen region.

Once we have filtered out the relevant body part, we can create a bounding box around the contours of the relevant area to calculate its length. We then tighten the bounding box to counter the loose clothing silhouette and also adjust it around the waist by using hip and knee coordinates.

We find the pixel length of the user’s height by marking the difference between the top of the head coordinate and the bottom of the toe coordinate.

We calculate the pixel-to-inch ratio by dividing the pixel length by the actual height in inches – which represents how many pixels are represented by one inch.

We now find the pixel length of the bounding box by identifying the difference between the left and right coordinates of the rectangle.

We convert the lengths of the bounding box into inches by dividing pixel length by pixel/inch ratio.

We use the length of the bounding box in inches as the parameter to the circumference formula and finally get the waist measurement.

Find out how Royal Cyber helped its client improve customer retention with analytics dashboard in this case study “Reducing the Customer Churn Rate for a Telecom Client.”

The user interface is a mobile application made for android. We use the simplistic and minimalistic UI to explain the use case of virtual body measurements by using a smartphone camera to click images. The system has been designed to transfer the images from the mobile app to the backend for processing and fetching the results.

To obtain their body measurements, all a customer has to do is open the application on their mobile device, enter their height in inches in the given space, take full body pictures in a standing position from the front and the side, and then simply wait for the results. Once the processing is complete, the user will be able to obtain their key measurements, including height, chest, waist, hips, and inseam.

Our solution seamlessly combines clothing-size measurement with the online shopping process to enrich the customer experience.

• Our model has shown a 90-95% accuracy with all the predicted measurements. With the help of this application, users can enjoy online shopping without having to worry about whether they are getting the correct sizes and doubting their decisions.
• The solution can also help eCommerce businesses by increasing customer engagement. The right decisions will lead the customers to make more purchases and result in high conversion rates.
• Our body measurement app eliminates the need to visit clothing stores in person to make “safer” purchases.

Our POC proves how we can use Artificial Intelligence and refined tech to design innovative solutions that benefit businesses and customers alike. Royal Cyber data science experts have extensive experience in developing customized solutions that meet the unique needs of businesses thriving in diverse industries, including health, manufacturing, fintech, oil and gas, and security. If you have any queries on the topic, feel free to contact us.