Refused Warriors

Colombo | The Art in Our Worlds

Martian Poster Generator

High-Level Project Summary

Martian Poster Generator is a project developed to illustrate scientific data in an artistic manner. Our web app consists of a machine learning model which can recolor greyscale images, a database to keep track of search inputs and their imagery, and a website at the front end to display results. This application takes user inputs as short texts and gets data from NASA APIs to generate informative posters which may boost the interest of general audiences who do not have a scientific background. As scientific data are presented in a more complex manner this application attempts to deliver it in a creative way.

Link to Final Project

https://github.com/SA-Tester/Refused-Warriors-22

Link to Project "Demo"

https://www.canva.com/design/DAFNzWY_Uug/h_0nMMdzow9SYpcNv-aP_w/view?utm_content=DAFNzWY_Uug&utm_campaign=designshare&utm_medium=link2&utm_source=sharebutton

Detailed Project Description

Image 1: Base Interfaces

What exactly does it do?

Recolour greyscale images
Generates word clouds for mission names.
Generates creative posters which can be shared on social media.

How does it work?

Our application is mainly based on scientific data related to Mars. The data gathered from the Opportunity, Spirit, and Curiosity Mars missions are used for our application(These data are received from NASA API " Mars Rover Photos" maintained by Chris Cerami).

In this application, we take user inputs such as rover camera type, sol, rover name, date(in earth format), and status of the mission. (The prototype project currently takes user inputs only for rover camera type and rover name ).

Based on the user inputs a query is generated, passed to the API, and data for three random images are parsed and stored in a MySQL database. These three images are also saved locally. The user input along with the IDs of the three generated images are stored in a table named "search_image" which helps to map possible user inputs and imagery later.

The generated images are then passed through a machine learning model which was developed using tensor flow and trained on an image set of 2187 images to recolour the greyscale images which are resized to be 256 x256 pixels. This model contains a deep learning neural network with twelve layers.

Then we use the Python PIL library to create a poster in 1920x1080 resolution using the above-recoloured images and data gathered from the API. (The prototype project is only employed with a template).To provide the correct credits we included the sources of information in the poster too,

Additionally, we created a word cloud generator using Python newspaper and WordCloud libraries for rover name user inputs. Since a better word cloud requires articles that have more than 1000 words, generating word clouds for rover camera name inputs is not effective.

Image 2: Generated Poster 1 for User Input: "opportunity"

Image 3: Generated Poster 2 for User Input: "opportunity"

Image 4: Generated Poster for User Input "PANCAM"

What benefits does it have?

Complex scientific data are illustrated in a pictorial manner.
Simplicity of presentation
Easy to share on any platform
Customizability

What do you hope to achieve?

To give a better understanding to the general public who lack a deep knowledge in scientific areas and boost their interest and curiosity in the field of science.

What tools, coding languages, hardware, or software did you use to develop your project?

Flask- Integrate the front end and the back end of the application.
TensorFlow-Develop the image recolouring neural network.
MySQL- Data storage
Python -Backend development

Python modules used:

Numpy module- Format images
Newspaper module-Download articles for the word cloud
WordCloud-Generate word clouds
Matplotlib & PIL- Image handeling
Skimage-RGB to LAB colour conversion

**The neural network was trained on a 2.21GHz CPU for 5 hours.

Space Agency Data

What space agency data you used in your project?

Mars Rover Photos API from NASA's Open API data portal to get the images required for our project. (Link to NASA Open APIs (Mars Rover Photos) - https://api.nasa.gov/)
Data from mars. nasa.gov to find descriptions of rover cameras

Links:

https://mars.nasa.gov/msl/spacecraft/instruments/chemcam/

https://mars.nasa.gov/msl/spacecraft/instruments/mahli/

https://mars.nasa.gov/msl/spacecraft/instruments/mardi/

https://mars.nasa.gov/msl/spacecraft/instruments/mastcam/

https://mars.nasa.gov/mer/mission/instruments/mini-tes/

How you used it?

Mars Rover Photos API -Get necessary imagery for the posters

Data from mars. nasa.gov to find descriptions of rover cameras

How it inspired your project?

These space agency data consist of a huge collection of information that can give a very clear detailed picture of any topic that we try to learn about. But to use the data, one has to search each and everything separately. This problem and the collection of data inspired us to filter and gather all the relevant data regarding a certain topic into one place so that when someone tries to learn about it they don't have to search separately.

Also, the data presented in Grayscale triggered the idea of image recolouring to us.

Hackathon Journey

HOW WOULD YOU DESCRIBE YOUR SPACE APP EXPERIENCE?

We had no prior experience in the field of ML and AI. The only perk we had was a strong determination and a clear target. Once the challenge summaries were released we discussed working on the challenge "Art In Our Worlds", and the idea never changed even at the most challenging times.

We developed the software right from scratch and we managed to learn everything that was needed along our journey. We can say that we went from nothing to everything. Learning all of such things has now become our passion.

From the beginning of this space app journey, our team “Refused Warriors” was able to overcome such challenges as the knowledge gap always through teamwork and dedication.

WHAT DID YOU LEARN?

We learned everything from hard skills to soft skills including using open-source APIs, Python web development, developing ML models, training them and using them in applications, teamwork, time management, presentation skills and much more.

WHAT INSPIRED OUR TEAM?

Our curiosity toward science and astronomy as well as our love for art became our inspiration. We were always fascinated by science fiction movies and apps that use visual images and graphics to spread scientific knowledge to the public in creative ways. And our passion to try new things prompted us to test and apply what we know in the field to bring a solution to the problem of presenting scientific data in an artistic and more accessible way.

HOW DID YOUR TEAM RESOLVE SETBACKS AND CHALLENGES?

The knowledge gap, having no access to resources or resource personnel, having no access to high-end computers and related components and even the current situation in our country challenged us more than ever. Unlike any other time

finding the time we need to learn and develop the software challenged us greatly. But, perseverance and good team support helped us to navigate through the hardships and reach our goal.

As our motto says, "Don't stop until you get accepted".

IS THERE ANYONE YOU'D LIKE TO THANK AND WHY?

Our heartfelt gratitude goes to NASA for organizing this competition on behalf of all the space lovers. Nonetheless, we are also very much grateful to the ‘SEDS’ Sri Lanka organizing committee of the Colombo division, and to all our mentors and colleagues for all the support thus, rendered. We would also love to thank our family members for providing us with the resources and motivation through the tough times. A big thank you goes to our team members who committed their study time to making this project a success.

References

https://api.nasa.gov/

https://mars.nasa.gov/msl/spacecraft/instruments/chemcam/

https://en.wikipedia.org/wiki/Hazcam