At the age of ever replenishing data with constant inflow, it is very tedious to classify astronomical objects manually which was the conventional way for a long time. The universe is full of stars clustered together to form galaxies. The study and analysis of these galaxies provides us with a vast understanding of our own galaxy. A deep convolutional neural network architecture for galaxy morphology classification and feature prediction is proposed. The galaxy can be divided into ten morphological classes based on its characteristics. The proposed architecture consists of two models: one for class prediction and the other for galaxy features prediction. This paper attempts to classify astronomical objects into Galaxies, Stars and Quasars using data from SDSS, LAMOST Surveys in addition to spectroscopic data analysis. Baseline approaches are compared for the classification task and improved optimised classification is further explored.