A Real-Time Object Dimension Estimation System

This paper presents a novel approach for real-time object dimension estimation systems. In order to create a trustworthy dimension detection system and investigate its applicability in many sectors, this study provides a novel approach for real-time object dimension estimation systems. In numerous disciplines, including manufacturing and engineering, architecture and construction, robotics and automation, augmented reality, and many others, the precise estimation of item dimensions is a crucial component. In this model, the realtime size feature is computed using the YoloV5 model. The proposed system is implemented on the Jetson. Nano-embedded development system. A custom dataset of 2400 images of each of the three classes (bottle, cup, and grocery bag) is used in the training of models. The detection and localization are performed using YoloV5. Recognition and identifying the images are done by Convolutional neural network (CNN) algorithms. The bounding box coordinates are used to estimate an object's size from a distance of 75 cm from the camera by using the Linear Regression method. The accuracy of the system's detection of the bottle, cup, and grocery bag is 88%, 84%, and 87%, respectively. The object dimension estimation has observed close to -0.1/ +0.1cm variation and presented a justified approach for the difference in accuracy evaluation of the detection and size estimation model.