|
Acadêmico(a): William Maurício Glück |
Título: ASTROLAB: UMA FERRAMENTA PARA IDENTIFICAÇÃO E CLASSIFICAÇÃO DE CORPOS CELESTES |
|
Abstract: |
Galaxy morphological classification is essential for astronomy. It allows a greater understanding of the origin and evolution of the universe. This work describes the development of a tool that performs morphological classification of galaxies using convolutional neural networks. The used design has two convolution layers, two max pooling layers, a fully dense connected layer using dropout and a output layer with softmax. The database used was created from the Sloan Digital Sky Survey (SDSS) and data provided by the project Galaxy Zoo 2. It contains 107,620 images for training and 978 test images. It was needed the creation of a processing step to reduce noise in the images used. The tool was developed in the Python programming language together with libraries to perform image processing and the creation of the neural network. Tests show a 86.40% success rate on the test dataset and 97.5% at the training dataset. |
|
|