Title: A catalog of broad morphology of Pan-STARRS galaxies based on deep learning

Abstract: Autonomous digital sky surveys such as Pan-STARRS have the ability to image a very large number of galactic and extra-galactic objects, and the large and complex nature of the image data reinforces the use of automation. Here we describe the design and implementation of a data analysis process for automatic broad morphology annotation of galaxies, and applied it to the data of Pan-STARRS DR1. The process is based on filters followed by a two-step convolutional neural network (CNN) classification. Training samples are generated by using an augmented and balanced set of manually classified galaxies. Results are evaluated for accuracy by comparison to the annotation of Pan-STARRS included in a previous broad morphology catalog of SDSS galaxies. Our analysis shows that a CNN combined with several filters is an effective approach for annotating the galaxies and removing unclean images. The catalog contains morphology labels for 1,662,190 galaxies with ~95% accuracy. The accuracy can be further improved by selecting labels above certain confidence thresholds. The catalog is publicly available.