Using Machine Learning in Refrigerator to Keep Inventory

Abstract

Intelligent Systems are emerging with multiple capabilities in day to day life. This is due to the advancing technology and rapid use of internet. Kitchen is area where intelligent appliances are used. There has been the advent of “smart refrigerators” but there is no way to make the traditional old refrigerators into a smart one. The intelligent system used in currently available products are expensive and also the user has to buy the whole system with refrigerator. Any existing refrigerator can be transformed into intelligent appliance using the Smart Refrigerator system. This paper documents how we have used machine learning in traditional refrigerators to make them keep inventory of what is stored inside them. The Refrigerator is able to perceive and monitor the contents and is also able to alert the user about the scare products. It provides an easy “plug and play” method to implement smart functionalities in traditional refrigerators without having to buy a new one. The detection algorithms run on a Raspberry Pi Model 3B with a webcam attached to it. This was mainly possible due to the recent advances in detection algorithms like YOLO v3 and Tiny YOLO which make it possible to perform object detection on small computers like Raspberry Pi.