The classification of galaxies based on their morphology is instrumental to the understanding of galaxy formation and evolution. This, in addition to the ever-growing digital astronomical datasets, has motivated the application of advanced computer vision techniques, such as Deep Learning. But these proposals have not allowed us to have a single pipeline that replicates detection, segmentation and morphological classification of galaxies made by experts. The process has been performed either visually or through relied on semi-automated software, mainly SExtractor. We present the implementation of a automatic machine learning pipeline for detection, segmentation and morphological classification of galaxies. Model based on Deep Learning architecture: Mask R-CNN. This state-of-the-art model of Instance Segmentation also performs image segmentation at the pixel level, which is a recurrent need in the astronomical community. We achieve a Mean Average Precision (mAP) of 0.93 in morphological classification of Spiral or Elliptical galaxies.
-
Install all libraries in the requirements.txt file. In their respective versions.
-
Download the configuration files of the training and validation dataset. When loading them in the respective folders leave the following name for both files: via_region_data.json
-
Unzip the imageset folder of the dataset. Duplicate it and incorporate the respective configuration files.
-
Download the weights and leave them in the "models" folder.