In this paper, two novel deep learning architectures are proposed to solve the scene classification problem using aerial images. The results that we get using the two models that are constructed utilizing a part of ResNET50 pretrained model are investigated. In model evaluations, we used one of the largest open access dataset, i.e. NWPU-RESIS45 dataset, that contains 45 different categories and in total 31500 samples. 95.7% accuracy that we get with the developped models show a competitive performance with the state-of-the-art methods. The proposed approaches are advantageous because they present a base model that requires low processing power and memory compared to the existing approaches. Within the scope of the study, the contribution of the regulation of the categories in a hierarchical structure to the classification performance is also investigated.