The two main objectives of this projects are implementing CNNs and evolutionary algorithms to a dataset of images of galaxies in order to classify them.
The project aims to classify, based on morphology, four different types of galaxies using a CNN.
⦁ preprocessing of an existing dataset of 10000 images of galaxies using scikit-learn
⦁ training of a CNN implemented in tensorflow using 128 epochs
⦁ confusion matrix
⦁ data analysis
Uses the previous CNN model as feedback to train an evolutionary algorithm in order to create a black and white image of a galaxy starting from gaussian noise.
⦁ generation of binary images with some random noise
⦁ running a suited evolutionary algorithm using scikit learn
⦁ analysis of results
I couldn't upload the datased because it's too big. instead I provided the code and a pdf preview of the evolutionary algorithm. If you needed the full dataset, feel free to contact me at tullio.formisano@hotmail.com