Find the outfit of your dreams

Finding an outfit may seem like an easy task, with more and more brands existing, and fashion influencers giving lots of inspiration. However, those influencers don't always share their secrets, and finding a perfect piece of clothing without knowing where it comes from can be very frustrating. What The Perfect Fit offers their users is the solution to this : even if the brands are not tagged on a social media post, we enable them to find them anyway.

In short, their lives will be made easier : all they need to do is screenshot the post in question, upload it on our app, and it will identify the references that you are looking for, and redirect you to the webpages where the pieces of clothing can be found!

Our target customers are quite obviously people who have an interest in fashion, are active on social media and like to shop online. Therefore, we targetyoung adults or gen Z people who can easily understand the concept and use it within a click.

To keep up with their need for quickness, the project needed to be as simple and intuitive as possible.

The Perfect Fit uses artificial intelligence in order to make it possible. Indeed, the more pictures are taken and uploaded to the app, the best it recognizes pieces of clothing. But that is not it! When a reference isn't available anymore, it redirects the user on a page with a similar piece of clothing.

Here is an illustration of how the user experiences it once they have downloaded the image. All they have to do is click on the text (which is actually a link) to go to the reference's webpage :

However, if the app can't find the right reference, here is what happens. If they click on the link, they are redirected to another reference :

What is great about the app is that it uses machine learning, which means that the more photos are submitted to the app, the more precise it gets. Based on what it already knows, this technology recognizes certain images, and more precisely, their content. When it doesn't, however, new references are associated with the features the app extracts from new images. The app acts like a database, since it saves all known images forever, but most importantly, like a smart one, since it can learn new information from new images and improve from it. The model adapts, and gets better : the possibilities are infinite.

For this to become possible, we use a library called ml5, which makes it far easier to elaborate the required code to be given access to artificial intelligence and machine learning.

More precisely, amongst this library, we picked MobileNet, which is a model that was built specifically for image classification. Here is how to build the code needed for the app :

First, an html file must be created, including :

Then, a json file, as it follows :

Big tech companies have already developed applications like ours. But the big difference lies in the fact that the apps aren't specialized in finding clothes, which makes their names far less eloquent. Google Lens is the perfect example : as developed and interesting as it is, it can be used for everything, from clothes to finding which flower one is seeing. One needs to know the app has this fonctionality in order to choose it to recognize clothes, which everyone does not. The main difference, therefore, is that our app is dedicated to this and aims at a very specific audience, which makes it more efficient.

Another type of competitor is apps like LikeToKnowIt. Those allow to find references of outfits on Instagram, but only if the person who wore the outfit uses it to put the references in question, which avoids the very problem The Perfect Fit solves, making the two apps quite compatible in the end.

The one remaining question we have yet to answer is : how is the app profitable? It is quite simple : between the many references the app is able to re-direct the user in case the outfit they are precisely looking for doesn't exist anymore, some brands are privileged by the technology more than others if they pay a commission when a piece of clothing of theirs is purchased after the user sees it on the app. Additionally, targeted ads, based on the previous researches the user made, appear from time to time on the app.

A project by Pauline Paris