Performance Analysis of TensorFlow2 Object Detection API Models for Engineering Site Surveillance Applications

Object detection in camera images is a powerful tool that can be utilized for a variety of applications. The COCO 2017 dataset assembled by Microsoft is one of the most extensive datasets publicly available for training and testing object detection models. The TensorFlow2 Object Detection API developed by Google contains a collection of pre-trained models using the COCO dataset and provides the tools for testing and deploying object detection models. Six of these models were used to test detections for motion capture images from SitePro Inc.