| Abstract | In the present context, there is one big issue regarding energy conservation. In fact, it is a valid question-why do we need to conserve energy? It is not at all wrong if the answer would be to save money. But certainly, it is not the only reason. Less electricity uses leads to less burning of fossil fuels. Although it is possible to generate electricity from the renewable and clean natural resources like water, wind, and sun, most of the electricity is generated by burning fossil fuels like coals and oil which are non-renewable natural resources and cannot be replenished once they get used up. They are not clean sources of energy either and are responsible for emitting greenhouse gases as well as causing pollution. Energy conservation also helps to prevent the threats to the ecosystem caused during procurement of coal and oil spills. There is an effort to reduce fossil fuel consumption by the means of nuclear reaction. Though it can generate a large amount of electricity, yet it is not the best solution as it produces radioactive waste products. Thus, decreasing the demand for electricity is very essential. To increase energy efficiency, the home automation system can play a significant role. The lighting system, as well as other electric appliances, could be made to automatically turn off whenever they are not in use even if the user forgets to switch them off. Earlier such systems have been implemented using various approaches in which specific sensor is used to measure a specific parameter. Use of those sensors make the circuit more complex and it may always not be reliable. This led to the idea of using a camera as a sensor. Using video surveillance, the changes in the captured area is measured as a result of which we can perform the desired automation. It is much reliable than the use of other sensors. Computer vision has made it possible to acquire, process, analyze and extract high-level understanding for digital images and videos. Computers can capture and process the images with details far better than human beings. In this study, YOLOv3 is implemented as an object detection system for human detection and finding if the appliances are on or off. This system replaces the use of predominant sensors by computer vision. With the advanced tracking by detection technique- Deep SORT, the movement of human beings is constantly tracked in order to increase the reliability of the system. Also, the modification of person re-identification model is presented to improve the performance of Deep SORT tracking. |