
The Astonish Computer Training behind the art -
Artificial Intelligence
What is AI?
Artificial intelligence is the name given to the process in which the computer makes decisions mimicking a human
In our case, we made computer to "create & design" like human
Machine learning
Machine learning is a more specific category in AI.
A definition for ML is "Made Model Based On Data"
Model & Data are the 2 keys in ML
Deep learning
Deep Learning is a more specific genre under ML, in this case, is the technique we use to create these beautiful Martian Artworks.
DL - Applies to every ML that uses Neural Network (an algorithm in ML)
Neural networks hold the name that made it sounds like the body part of human, in fact, their functions are similar!
3 WAY APPROACHES
TURNING STEM INTO STEAM
All in Machine learning!
Projected GAN
These beautiful Martian landscape images are all generated by Projected GAN (Generic Adversarial Network), an algorithm under Machine Learning.
The database for training is HiRISE images from NASA's HiRISE project.
For the training, we take in HiRISE images taken by power cameras and cut them into 512 x 512-pixel images, the "real" images will be collected as input for the computer, and generate images output.
​

Stable Diffusion
Stable Diffusion is used in generate images based on prompts ​
Prompts related to Martian Colonization and Mars Exploration gave us a good impression of how computers can do art and their unique creations.
​
We write Python code connect with HuggingFace.co in running the Stable Diffusion output. For prompts, it can be anything as general to "Star Wars" to "An astronaut playing with rabbits on Mars".
Style Transfer
How do combine science & culture? The technique of Style Transfer gives us an answer.
​
Style Transfer is performed when 1 content image + 1 style image adds together. In a simple way, the style image gives a "style" to the content, a new image is created.
​
In our case, we use Martian images as style, when it gives a pattern to the Egyptian clay vase, new style booms.


