Virtual try-on

A feature for the main Style app that helps you to try on clothes from the comfort of your home.

Comment 2

Style

App for virtual fitting of clothing.

The project was created as a feature for our main application for generating fashionable looks from existing items based on your preferences and global trends.

Since many countries are quarantined because of COVID-19, people can't go shopping like they used to. So they are in need of alternatives. We have developed an app where a person can try on clothes without leaving home! Now you can check whether the pretty black T-shirt from the online shop will fit your favourite jeans!

Main functionality:

This app is designed to show you how you will look in clothes you select from the catalog.

Clothes\Persons Person1 Person2 Person3 Person4
Cloth1 Person1_0 Person2_0 Person3_0 Person4_0
Cloth1 Person1_1 Person2_1 Person3_1 Person4_1
Cloth1 Person1_2 Person2_2 Person3_2 Person4_2

How does this happen:

  1. You choose the type of clothing (t-shirt, pants, sweater, etc.).
  2. You choose a specific clothing model.
  3. Upload a photo of a person.
  4. Get a generated photo of this person in the selected clothing.

Implementation

The app consists of two main parts: an Android app and a REST API server.

Android

Withe the help of Android app user selects things to try on and uploads their photo. Then the app sends data to the server and receives the image that has already been generated.

Technologies used: Android Studio, Java, and OkHttp.

Server

The server processes the image. Communication with the client-side is performed via the REST API. Several models are implemented on this side: segmentation of clothing, segmentation of the human head and body, definition of the key points, and a model for image generation (GAN).

Technologies used: Flask, Python, PyTorch, Detectron2, DensePose, and Mask R-CNN.

Built With

Share this project:

Updates