Deep Learning in Construction: Review of Applications and Potential Avenues

Jacobsen, Emil L.; Teizer, Jochen

doi:10.1061/(asce)cp.1943-5487.0001010

Cited by 17 publications

(8 citation statements)

References 71 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Research within computer vision has seen an increase in interest from construction research (Jacobsen and Teizer, 2022). The methods based on images and videos have also been used for activity recognition.…”

Section: Computer Vision Methodsmentioning

confidence: 99%

“…This kernel can be seen as a filter running over the input, filtering the data to find features. In stacked convolutional networks, the deeper layers will represent the features in more abstract ways, which have significantly impacted language processing (Krizhevsky et al, 2012) and computer vision (Jacobsen and Teizer, 2022). The method of combining convolutional layers with LSTM also has shown robust performance on human activity recognition before (Ordóñez and Roggen, 2016).…”

Section: Time Series Classification Modelmentioning

confidence: 99%

“…Work sampling can be used to estimate CLP due to the direct correlation between the direct work (DW) category "producing" and productivity. Automated processes, especially machine learning methods, have also been investigated to estimate productivity using static information such as project size, worker experience, or contractual agreements (Jacobsen and Teizer, 2022). The methods differ in the modality used for activity recognition, and most publications fall into one of three categories: images/videos, audio, or kinematic data (Sherafat et al, 2020).…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Automating the Estimation of Productivity Metrics for Construction Workers Using Deep Learning and Kinematics

Jacobsen¹,

Teizer²

2022

Proceedings of the 29th EG-ICE International Workshop on Intelligent Computing in Engineering

Self Cite

View full text Add to dashboard Cite

In this study, a novel method for direct work estimation is used to classify whether a painteris performingdirect workor not.The aim is to build an accurate and reliable work classification algorithmthatcanhelp monitorconstruction sites.The method utilizes adeep learning algorithmusing convolutional and long short-term memory layersto classify multivariate time-series data collected from five inertial measurement units (IMUs)mounted on the workers’arms, torso,and legs.Three models are developed, differing in window sizes from 3 seconds to 7 seconds.Thebest performingmodel achieves an accuracy of 90%and anF1-score of 87.6%. This is thefirst step towards a general model that canclassify productivity measures for workers on construction sites, which will be a valuableinput formonitoringconstruction sitesandfuture analyses.

show abstract

Section: Computer Vision Methodsmentioning

confidence: 99%

Section: Time Series Classification Modelmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Automating the Estimation of Productivity Metrics for Construction Workers Using Deep Learning and Kinematics

Jacobsen¹,

Teizer²

2022

Proceedings of the 29th EG-ICE International Workshop on Intelligent Computing in Engineering

Self Cite

View full text Add to dashboard Cite

show abstract

“…The literature review and the bibliometric analysis are based on the Scopus and Web of Science (WoS) scientific databases because of the extensive coverage of literature, the ability to export the search results as comma separated values (CSV) file for further analysis, and the support of Boolean (i.e., "AND", "OR") and proximity operators (i.e., "W/", "NEAR/") in search strings for advanced queries. The selected scientific databases have been successfully used in previous state-of-the-art review papers (Jacobsen and Teizer, 2022;. The yielded search results have been exported in CSV file format and used for further analysis.…”

Section: Methodsmentioning

confidence: 99%

“…Three categories of input data were found to be used for trajectory prediction in construction literature: vision-based data, raw location tracking data, and 3-dimensional point cloud data from LiDAR sensors. Similar to Jacobsen and Teizer (2022) the average time of publication (ATP) is used as a bibliographic metric. Table 1 shows the ATP for the different types of input data, prediction methods and applications in the construction literature.…”

Section: Trajectory Prediction In Constructionmentioning

confidence: 99%

Trajectory Prediction: A Review of Methods and Challenges in Construction Safety

Chronopoulos¹,

Teizer²,

Esterle³

2022

Proceedings of the 29th EG-ICE International Workshop on Intelligent Computing in Engineering

Self Cite

View full text Add to dashboard Cite

This review focuses on methods for trajectory prediction of moving entities (i.e., pedestrian workers and heavy construction equipment) in construction. To the authors’ knowledge it is the first review on trajectory prediction devoted to construction safety. Through a bibliometric analysis of the relevant literature, it examines the input data and prediction models used for trajectory prediction in dynamic and complex construction environments. Several techniques are available to perform prediction and their performance varies widely. Various types of data is being used, however, so far v ision-based data is the major input to the models. Hence, computer-vision techniques are deployed for object tracking to infer the locations of the construction resources in almost entirely outdoor environments. This review concludes with an overview of the gaps, challenges, and future research steps for trajectory prediction relevant for researchers as well as practitioners working on reducing occupational health and safety hazards on construction sites.

show abstract

Development of an action classification method for construction sites combining pose assessment and object proximity evaluation

Kikuta,

Chun

2024

J Ambient Intell Human Comput

View full text Add to dashboard Cite

Addressing the inherent hazards of on-site construction work and stagnant labor productivity is crucial in the construction industry. To tackle these challenges, automated monitoring of construction sites and analysis of workers' actions play a pivotal role. In this study, we developed a method for classifying actions at a construction site from video, using deep learning. Specifically, we used two image processing techniques, pose assessment and object detection, and found that the accuracy of action classification was improved by extracting information on the proximity of workers to equipment installed at the construction site, and also by considering the pose information. For classification, LSTM (Long Short-Term Memory), CNN (Convolutional Neural Network), and XGBoost models were used, and the presence of proximity information improved average recall by 7.0% to 8.5% for all models used. The final model was developed as an ensemble of these methods, offering accuracy and average recall that are higher than with conventional methods. The methodology developed in this research enables quantification and visualization of work content at construction sites, contributing to the overall enhancement of safety and productivity within the construction industry.

show abstract

Deep Learning in Construction: Review of Applications and Potential Avenues

Cited by 17 publications

References 71 publications

Automating the Estimation of Productivity Metrics for Construction Workers Using Deep Learning and Kinematics

Automating the Estimation of Productivity Metrics for Construction Workers Using Deep Learning and Kinematics

Trajectory Prediction: A Review of Methods and Challenges in Construction Safety

Development of an action classification method for construction sites combining pose assessment and object proximity evaluation

Contact Info

Product

Resources

About