Visual Machine Intelligence for Home Automation

In this paper, we present a vision-based machine intelligence system to sense the ON/OFF state of commonly used household appliances like Television, Fan, and Tube light/CFL. The visual modality of sensing the state of appliance acts as a proactive substitute for sensors used in the home automation network. The sensor-based home automation system suffers significant complexity in terms of cost and implementation whenever the suite of devices/appliances in the home automation increases. To overcome the complexity of home automation system in terms of cost and implementation, our proposed system uses vision-based machine intelligence to detect the states of multiple appliances without being dependent on the number of appliances. This proposed method of sensing the states of appliances thus give rise to a novel home automation system. The proposed home automation interface has multiple advantages over traditional home automation interfaces which incorporate the use of various discrete sensors. A switch to toggle between the visual monitoring of the home and sense of the home without sending images makes this system a bimodal approach towards Home automation which keeps a check on the privacy issues. The proposed machine learning based approach was found to effectively switch ON/OFF the appliances like Television, Fan, and Light and accordingly, update the user about the state of the appliance on a website.