FashionSwap: Image-Based Clothing Cropping for Style Transformation

${\textsf{\color{magenta}FashionSwap}}$ is an innovative AI project that focuses on the transformation of clothing styles. Using a dataset of individuals wearing different clothing items, the system segments or detects the clothing item, crops it, transfers its style through a GAN, and finally fits it back onto the individual. This way, we can visualize how a transformed clothing style would look on a person.

Team Members

Fabian Perez , Paula Arguello , Miguel Molina

Objective

Transform clothing into images of people by segmenting, cropping and transferring style using AI

Datasets

We've utilized the following datasets in our project:

1. ${\textsf{\color{violet}Fashion Controlnet Dataset V3}}$ Another valuable dataset from HuggingFace which has been essential for our project.

   • Access the dataset here

2. ${\textsf{\color{violet}Texture Dataset}}$: Texture dataset made by hand, with a set of different types of textures

Models

Neural style transfer and Unet

Links

1. Slides
2. Video