Smart Refrigerator Management System Using Deep Learning

In today’s modern era smart home appliances with new features and technologies plays a vital role to make human life easier. One of the essential and common home appliances is a Refrigerator which helps to preserve Food and Vegetables. There are many Smart Refrigerators available in the Market with integrated cameras WI-FI and Embedded Technologies. By incorporating artificial intelligence and deep learning to enhance the refrigerator’s functioning, smart refrigerators aid in the management of food. Deep Learning Technology is being used to detect the fruits inside the Refrigerator. This is necessary to detect the fruits present inside the refrigerator and keep the track of count present inside it productively and precisely. The goal of this work is to optimize and deploy the trained model in order to capture and categorize images inside the refrigerator in real time. To make fruit classification and recognition more accurate, we created an customised dataset to train and test the model with YOLOv7 deep learning algorithm. Proposed work is also extended to classify the bad and good quality of fruits. The entire system has been implemented using hardware based on NVIDIA Jetson Nano. The suggested system employs a wide-angle camera to capture images towards object recognition and inventory control. It provides an augmented pictures by considering different positions, perspectives, and lighting conditions along with manually created labels. This system also improves the occlusion problem of fruits present inside the inventory. Evaluation of this work reveals that the accuracy is obtained up to 98% with 35 FPS against the YOLOv3 which is attaining an avarage accuracy of 92.45%.