Inspiration
We've all watched a movie, a series, or just seen someone random on the street and said to ourselves "Damn, their outfit rocks!". And then we might have searched throughout stores for those pieces of clothing that look similar but there could have been the following outcomes: 1. time wasted because we did not find anything, 2. found something slightly similar and accepted that there cannot be a perfect match, or 3. we actually found something similar. But sometimes, the stores we end up looking in are not necessarily the most qualitative in terms of textiles.
What it does
Hence, we thought about an app that could find the outfits either from gallery pictures (screenshots, savings) or by taking a picture with a camera and linking the findings to the Inditex database (which is known for having quality materials in clothing). This app would find the match between the picture and what Inditex has to offer (for example a match between the photo and Zara) and the customer could either save it in a personal collection (in case they want to buy it in the future) or could be redirected to the shop cart of the store and buy it. The app has a built-in user profile that would remember the personal measures of the customer and their preferences (height, size, season, section), to get the match as accurately as possible. Moreover, this idea could also be transformed into a web extension with the same properties when searching on a PC.
How we built it
In order to build a project we needed to build a team, and that is what we did during the team-building. In terms of the project, we started designing the product view and attempting to download the entire database provided by Inditex. Afterward, we initiated a GitHub repository and developed both mobile and web version frontend. Additionally, we used tools as Google Colab and Jupyter Notebook for AI training. Finally, we have been working on connecting these two parts by constructing an API for image transmission and responses but we couldn't finish it :cry:
Challenges we ran into
At the outset, downloading the entire database was truely challenging due to its sheer magnitude. Furthermore, we encountered numerous hurdles while attempting to access the server repeatedly within a tight timeframe. However, with the assistance of Inditex representatives, leveraging the exceptional ChatGPT , and investing countless hours, we devised a solution to streamline image downloads using thread concurrency, eliminating prolonged wait times. We managed to download 14.000 pictures in one hour. Another daunting task was developing the entire frontend, compounded by the fact that not all of us had prior experience in this domain. Moreover, training the AI and attaining satisfactory outcomes proved to be an arduous endeavor, demanding significant focus. Nonetheless, we ended up making a good model of determining similarities between the pieces of clothing.
Accomplishments that we're proud of
We are delighted to have reached our goal and completed the project on time, especially from an algorithmic perspective. Additionally, we take pride in the aesthetic appeal of our design, as the app closely resembles the initial design concept. We are proud of our team's ability to learn and adapt, effectively addressing unexpected challenges as they arose.
What we learned
In my humble opinion, hackathons are not merely competitions, but rather environments where one can let their imagination soar and collaborate on truly innovative projects with like-minded individuals. These events also feature experienced mentors who are eager to support your growth and encourage continuous learning. Personally, I've learned a great deal from my teammates, as well as from friends I've made during meals and social activities. The goal is to expand your network and step out of your comfort zone to foster personal growth. I believe everyone has gained new skills, whether it's mastering a programming language, learning to use a new platform or IDE, or experimenting with hardware devices.
What's next for Matcher
This idea could go even further- by relying on fashion icons and fandom (such as F.R.I.E.N.D.S, Taylor Swift, Harry Styles), we could integrate a search bar based on natural language so the customers could search for the outfits based on the characters they want to dress like. As the fashion stores not only provide clothes, we would also like to focus on accessories-, shoes-, perfumery- or decorations- detection. Lastly, with this comprehensive ecosystem of Matcher, we would provide the customers not only with a product, but also with a new kind of experience.
Log in or sign up for Devpost to join the conversation.