FitFinder

Main Page
Upload A Photo
Item Found
Shopping Cart

Inspiration

The idea for FitFinder came from the frustration of seeing an outfit on someone and not being able to easily find and purchase the same or similar items. We drew inspiration from other popular machine learning apps such as Shazam, which identifies the song name from music. We chose to do a mobile app, since there has been a strong increase in usage of social media and online shopping through mobile phones.

What it does

FitFinder is a mobile-first web app that allows users to upload an image of someone wearing an outfit and it will find links to buy the same or similar clothing piece, and include details such as the price, sizing, colour options, brand name, and model photos. We also include a list where users can save their favourites, and add each piece to the cart.

How we built it

FitFinder was built using a combination of technologies including a website built in React, prototyping and design in Figma, and machine learning technologies such asPython, TensorFlow and Keras for image recognition and object detection to identify the clothing items in the image.

Challenges we ran into

One of the main challenges we faced was the time crunch of the hackathon. Due to the limited time, we were not able to implement every single page of the app that we wanted. Additionally, some parts of the system had to be put aside in order to meet the deadline. We originally attempted to use a machine learning method for finding outfits, but we encountered difficulties in obtaining sufficient and reliable data sources. As a result, we shifted our focus to using a similarity algorithm with an existing list of images.

Accomplishments that we're proud of

Despite the time constraints and challenges we faced, we are proud of being able to successfully implement a functional prototype of FitFinder. We were able to gain experience with machine learning techniques and technologies, which was a core goal of our app. We are also proud of the ability to find a way to achieve our goal, even when the initial approach failed.

What we learned

Throughout this project, we learned the importance of time management and the challenges of working under time constraints. We also learned about the challenges of obtaining sufficient and reliable data for machine learning models and the importance of having a backup plan. Additionally, we gained experience in implementing image recognition and search algorithms, as well as the difficulties in trying to match real-world clothing items with products sold online. Furthermore, we have learned the importance of focusing on the most important features and trying to get a MVP working.

What's next for FitFinder

In the future, we hope to improve the accuracy of the image recognition and object detection, as well as expanding the number of e-commerce websites that the app can crawl. Additionally, we plan to add more features such as the ability to save favorite outfits and receive personalized outfit recommendations.