SizeMe: eCommerce

Figma Wireframe of SizeMe

Inspiration

Our project idea stemmed from a problem in e-commerce, especially in clothing and fast fashion. Why do all shirts, or pants that are technically the same size not fit as such when purchasing from a host of online stores? The resulting process of filing for a return, handling the reimbursements is both painstaking and intensive from both a client-side and for the manufacturer.

What it does

In comes, SizeMe => a simple web application which uses computer vision and machine learning to decipher your body measurements, and provide the corresponding size catered to each manufacturer. Using SizeMe is a simple, three step process: upload a picture where your body is in the frame and wait for the system to output the corresponding measurements. From these measurements, the app will compare it with the manufacturer suggested sizing chart to provide a recommendation best suited for your body type. Then, the user has the option to directly proceed to apparel offering with the right size selected (where possible).

How we built it

Backend: Flask & CNN (Convolutional Neural Networks) for image analysis Frontend: Bootstrap

Challenges we ran into

Faulty image detection when we first tested the model. Inability to integrate a functional OpenCV model into this web application, for real time image processing.

Accomplishments that we're proud of

Training a large model to detect images, and derive the corresponding article measurements was still a major accomplishment. Given it was a first hackathon for many, we would have loved to overcome some of the hurdles above but time constraints did play a role.

What we learned

We learned the importance of prototyping from a front-end perspective, in order to deliver a compelling user interface. While our prototype was a significantly more vibrant design, we still believe we've put together an interactive web app.

From the back end's perspective, we learned the importance of giving adequate time towards CNNs to ensure robustness and reliability. From finding the dataset, priming the model and then integrating it into a fully functional web app, it was a lot to complete in 24 hours but we feel we learned the fundamentals, which we can carry with us to future competitions.

What's next for E-Commerce App

SizeMe is scalable, with great scope to further analytics-driven insights for consumer apparel brands on the types of clothing and sizes that are most popular for each season. In the future, SizeMe can become a one stop shop (using webscraping methods) to display articles of clothing which best suit each user's individual measurements and body types.