**1. Introduction**

Construction work is tedious and subject to delays, and its quality may be compromised by many factors, such as construction equipment, workers, and materials. Therefore, it is necessary to improve construction quality and progress in today's increasingly competitive market by considering good job site management and meeting construction costs. At a job site currently, a job site manager oversees everything construction-related, including workers, machines, and materials [1–5]. The manager has to take care of virtually everything at the job site [6]. The improvement of management methods using innovative technology helps to not only accelerate the development of the construction industry but also improve a company's competitiveness in the market.

Most general contractors deploy imaging devices, such as photo and video cameras, to document the progress of construction activities throughout the entire process. The image data are collected, in general, by filming with a mobile camera operated by a worker or a video camera set up at a fixed location. Most image data collected are used passively for reference or even just shelved. The others are used to prepare quality documents or demonstrate construction status and progress. Suppose artificial intelligence (AI) is introduced to recognize objects in the images and help job site management identify and tag things in the images. In that case, these image data may serve as an essential basis

**Citation:** Lung, L.-W.; Wang, Y.-R. Applying Deep Learning and Single Shot Detection in Construction Site Image Recognition. *Buildings* **2023**, *13*, 1074. https://doi.org/10.3390/ buildings13041074

Academic Editors: Maxim A. Dulebenets and Saeed Banihashemi

Received: 20 January 2023 Revised: 28 March 2023 Accepted: 9 April 2023 Published: 19 April 2023

**Copyright:** © 2023 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

for decision-making within construction activities, including construction planning and design, job site safety, automated equipment management [7–10], and quality monitoring and maintenance. For example, suppose a specific machine is tagged in video footage of construction activities [11]. In that case, the project team may exploit the captured data for project decisions of route management, machine setup, and site safety [12–15].

When recognizing and classifying objects in many images, a deep learning model may be introduced to accelerate the extraction of high-value digital information crucial for construction management. The mainstream in developing the deep neural network is the convolution neural network (CNN) which extracts critical feature information by including one or more convolution layers and pooling layers through a combination of algorithms and multi-layer computation of convolution neurons as the images are converted into data [16]. The feature information is fed to the neural network for training in a fully connected manner until identical or similar features in the same class of images are identified and documented. The relative locations and features digitally arranged during the recognition of new images are systematically computed and processed to identify the similarities between images for successful image judgment [17,18].

AI is having revolutionary impacts on construction engineering [19]. Thanks to the powerful capability of AI in data processing, analysis, and searching for massive digitization, a model to recognize construction objects at a job site can be built to rapidly and accurately identify workers [20–22], machines [23,24], and materials [25] in job site footage while tagging their relative locations in the images to provide more site-related information for project management, which is a rising topic in the industry in the pursuit of breakthroughs and innovation.
