High-Level Project Summary
Our web application brings the use of NASA's data to the next level.We created an intuitive, interactive and easy way to create art from any picture זand style.You can search for a picture from NASA's database and choose your favorite picture that inspires you to create an art with.By combining your personal pictures like an art from a known artist that inspires our generations or something special for each one of us you are creating something truly unique for you.Also it can be used to share positive social messages that bring people together.
Link to Final Project
Link to Project "Demo"
Detailed Project Description
We created a web application that can make art from search words, using images from NASA's Image and Video Library.
The search word "moon" for example, yields a choice of images from NASA's archive. The user can then choose the moon image he/she wants, and then upload or choose a style.
After that, the picture and the style are transferred to GANS algorithm that creates in a couple of seconds a new, creative and unique picture of the moon.
On the left is the original moon picture, on the middle is the creative moon, and on the right is the style that we used.
Abobe: VanGoch Earth. Below: Ukraine moon
Space Agency Data
We used NASA's API for Image and Video Library
Hackathon Journey
Introduction
We aim at building an app where the user writes a search expression (earth), and as a response, we display an artistic style image of the searched object.
Methods:
We use “NASA Image and Video Library” (API docs) and send a query in the following format:
url = 'https://images-api.nasa.gov/search?q=moon&media_type=image&page=1'
The query yields a list of images. We download a few and choose one to further manipulation.
We feed the image into a CycleGAN model which was trained on the VanGoch image. For “earth” input, the output is VanGoch style earth.
References
Examples of how to create new pictures from different styles:
- https://www.kaggle.com/code/duyminhle/pretrain-cyclegan-tutorial
- https://www.kaggle.com/code/ohseokkim/transfering-style
Tags
#tensorflow #gans #python #javascript #application #social justice