Mango Outfit Project

Inspiration

We think that AI together with Big data can help in many fields. One of them could be fashion. We could create AI models that learn about actual trends, and help the costumers in different situations, such as deciding their outfits for a date, easily finding the clothes that best pair with one clothe of their wardrove, and many more.

What it does

We have developed a system that, with a given product (from the Mango Dataset) generates some outfits with other clothes from that dataset that fit with the choosen one. Addictionally, we also use the Stable Diffusion AI (link) to generate an image with an outfit made with similar clothes.

How we built it

First, we had to analyze and then filter the Dataset from Mango, to keep the columns that we found important to train our machine learning model and remove some products that we found irrelevant for the outfits. Then we trained a gpt2 model with the outfits and some prompt engeneering to it learns the products from the database and the combinations. Finally we generate a image with a fake model wearing the clothes with a stable diffusion.

Challenges we ran into

Pre-process the dataset to identify the most important features for our system.

Accomplishments that we're proud of

Our system is capable of generating some outfits with the provided product and generate also an example image with a fake model with the complete outfit.

What we learned

How to use the pandas python library and work with .csv files with python scripts.

What's next for Mango Outfit Project

It would be nice to add a function where the input is an image of one product (that may not be from the Mango Dataset) and the program is capable of identifying the characteristics of that product to generate a suitable outfit with Mango products.