Real-Time Object Detection with Tensorflow Model Using Edge Computing Architecture

Abstract

This paper presents the capturing of objects using Wi-Fi enabled modular esp32 camera and processes the captured stream of data using machine learning and computer vision techniques, then sends the processed data to the cloud, there are major cloud providers in the market who occupied more than 80% of the global public market the cloud providers are Google Cloud, Amazon AWS, Microsoft Azure. Google Cloud Platform (GCP) is been our primary choice because of its good documentation availability, The Cloud IoT-Core Gateway, as well as a serverless cloud layer to store all of the data. The cloud functions help to trigger the notifications to the users when the cameras detect what we have trained the model. The Edge computing project uses an ESP32 With cameras as a device listener and a raspberry pi as an edge server which has an image classifier model trained with TensorFlow.