We develop an autonomous pipeline for morphological classification of Pan-STARRS galaxies. The process relies on fetching the galaxy images, applying a collection of simple filters, and classification by a deep convolutional neural network (DCNN) with LeNet-5 architecture. The process provided a catalog of 1,662,190 galaxies, labeled by their broad morphological type, with ~95% accuracy compared to previous catalogs. The classification accuracy can be further improved by selecting a subset with higher classification thresholds. The training data were prepared with manually classified debiased “superclean” Galaxy Zoo annotations of SDSS galaxies that were also imaged by Pan-STARRS. The training data were augmented by mirroring all Galaxy Zoo galaxies to avoid the possible bias from human perception. Our analysis shows that the DCNN approach is sufficient for generating high-accuracy labels and removing unclean images. The labeled data will be further analyzed to identify and profile specific feature of the galaxy arms such as the number of arms, shape of the arms, and more.