The mission of the smart city is to improve the infrastructure and services to resourcefully manage the growing urbanization, maintain a sustainable environment, and improve the economic and living standards of their citizens. Various growing fields of artificial intelligence are expected to perceptively support the sustainable development of smart cities. People living standards are improved by incorporating the technology in their daily activities to provide the growth of smart cities. This study reviews the theoretical perspective of how deep learning can be applied to the development of smart applications. A comprehensive insight has been brought into the deep learning algorithms involved in applications like waste management and the healthcare domain. Specifically, this paper discusses the significance of deep architectures to classify the waste images into recyclable or not. Additionally, the development of the smart imaging sector to diagnose diabetic retinopathy pathology has been addressed. This paper reviews the freely available datasets, extensively used pre-processing steps, and analyzing the performance of DL algorithms for the aforementioned applications. We also discuss future research directions where the DL techniques can play a significant part to realize the concept of intelligent applications.