Nowadays, the current research trends in smart buildings plays a vital role and acts as an integral part of smart grids and smart cities. Recently, the need for sustainable energy and efficient power consumption has become very important across the world especially in domestic and industry zones. The challenges of global environment and energy sustainability can be addressed by the current thrust research area of smart buildings. This paper presents a power conservation system that utilizes deep learning techniques to detect and monitor human presence in order to enhance energy utilization in various environments. The proposed system credits open-CV algorithms and machine learning models to precisely identify humans and their movements within a given area, enabling intelligent control of power-utilizing devices. The Internet of Things (IoT) and deep learning techniques are used to control the operation of lights and fans and also enable a security feature in the form of telegram bot. Experimental results emphasize that the deep learning-based recognition can detect the persons in a particular area with an accuracy of 96% and 99% in dark and bright areas conditions, respectively.