Object detection model of interior housing object for construction company

Abstract

With the advent of object detection models, datasets are being created and used almost every day. YoloV5, the current state-of-the-art model, has outperformed other machine learning models thus far and is being used in many different applications. It is also being used in the construction industry to ensure worker safety and used for the detection of construction materials.

However, as other models lack general processing time and affect the efficiency of the system, YoloV5 covers both aspects of performance and accuracy. This project proposes using the YoloV5 machine learning model for developing a mobile application relevant to the construction industry, specifically the interior housing and carpentry trade business. However, as interior housing datasets are not widely available, gathering interior housing images and annotating them is time-consuming and costly. We also research using the unity perception package as part of the unity compared to collected datasets in terms of its cost and effectiveness.

The project’s dataset would be evaluated using the “weights and bias” platform to analyze its prediction precision.