FIT-FASHION

Inspiration

Being able to buying clothes online but not being able to see yourself in them has been a major issue for anybody that wants to order clothing in the comfort of their home.

What it does

Thus by just taking a video around yourself you will be able to create a 3d model of yourself which you can fit any kind of clothing onto.

How we built it

We used Gaussian splatting to create the 3d model which was then fed into a fine tuned diffusion model that then creates multiple frames of the input image with the specific clothing that the users wants. These frames are then combined together with the Gaussian splat model to create a 3d model of the input model with the different clothing.

Challenges we ran into

Running the Gaussian splat model, getting the right data to train the diffusion model and the training loop to train the diffusion model were the hardest challenge we had to take care of.

Accomplishments that we're proud of

Building a pipeline to be able to create such a cool concept, finishing bug after bug and fighting till the end

What we learned

We learned how to implement computer vision models along with the mathematics they require.

What's next for FIT-FASHION

Creating a process of fitting the cloths on the model such that it will take account of the musculature and the body composition so we can accurately predict how the piece of clothing will fit onto the model