| Abstract [eng] |
In this thesis the populations of star clusters of two spiral galaxies were analyzed using convolutional neural networks for cluster detection and their parameter inference. A method based on the ResNet convolutional neural network was developed, which is suitable for deriving evolutionary (age, mass), structural (size), and environmental (extinction) parameters of star clusters, as well as localizing both semi-resolved and unresolved star clusters in multiband images. Mock star clusters in combination with real galaxy backgrounds were used to train the network. The method was tested both with mock and real star clusters in the context of the M31 and M83 galaxies, showing a good agreement with previous studies. A full pipeline for the detection of star clusters and their parameter inference has been developed in the context of the M83 galaxy. Using the method, 3,380 star cluster candidates were found and analysed with respect to the spiral arms and center of the galaxy. An age gradient with respect to the spiral arms was observed in agreement with the spiral arm density wave theory. In addition, more dense cluster candidates were found towards the galactic center and less in the outskirts of the galaxy. |
