Tree species classification is an active area of research, in forestry related tasks, which is quite challenging but has useful applications. It is difficult to meet the challenge due to complexity of the problem as well as due to non-availability of datasets for research purposes. There has been almost no work done in Pakistan in this area of research due to the earlier mentioned challenges. This thesis makes the following contributions: (1) dataset creation of 5 tree species consisting of 1311 images including leaves, trunk and canopy, (2) dataset containing a total of 305 full tree images of 5 tree species, (3) propose a complex model (parent-children) architecture taking three input types i.e., trunk, leaves and canopy in a single input example. The parent model contains three child models, i.e., one sub model each for trunk, leaves and canopy. Based on the input images, the model(parent-children) is trained as a whole and used to classify to a single output class (4) Furthermore, a single model which takes the whole tree as input for categorical classification is proposed as well. It is a novel piece of work for the species that exist in the homeland. The proposed models give very good results.