Skip to content
Build App Prototypes using the Power of Deep Learning Based AI
Jupyter Notebook Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.DS_Store
.gitignore
README.md
app.py
middleware.py
sc0.png
sc1.png
sc2.png
trainClassifier.ipynb
ui_detailsCategory.csv

README.md

copraML AppPrototyper

Build App Prototypes using the Power of Deep Learning Based AI

Winning hack at Hack2Innovate AI Hackathon NVIDIA challenge!

screen0

What if making an app prototype is as easy as selecting some images?

We have tried to solve exactly this based on a easy to use process, where the user selects the type of the application he would like to begin with, eg: Social or Travel app and then chooses a couple of UI screens and a basic colour palette. The webapp then makes predictions based on the selected screens and the colour palette to suggest them a UI prototype which is visually appealing and readily presentable

screen1

screen2

We use a rich dataset having more than 70,000 UI Screens from various apps on the play store. The Dataset can be found here : http://interactionmining.org/rico

We use the trained vectors from a Deep Convolutional Autoencoder to find semantically similar images to the ones selected by the users.

We also have custom classifiers for different Views for Android Apps - ListView, CardView and GridView. These provide the parameters like optimal size, position and colour for the UI screens. We use a custom Convolutional Neural Network roughly based on the VGG architechture.

This information is then fed to the templating system which builds a visually appealing prototype for the user

There is a lot of possibility for extension, and we have just scratched the surface of the possibilities in this Hackathon.

Thank you!

You can’t perform that action at this time.