Presenter(s)

Ishan Vijay Ghutake

Download

Download Project (15.5 MB)

Description

The construction industry is going through a huge shift toward automation, with safety being one of the major challenges. We always want to take measures through which more accidents resulting serious injuries and deaths could be avoided. Indeed the construction sites are bound with several safety rules, one of the most important is having required personal protective equipment (PPE) based on the worker working environment. The presence of the monitoring camera at construction site provides an opportunity to enforce these safety rules by applying computer vision techniques and algorithms. This study shows capability of the Deep Learning model to classify worker as safe and unsafe and provides logical explanation to strengthen the prediction result. Here we exemplified classification of worker by using five convolutional neural network models with various layer structures. We collect a dataset of construction site scenes and annotate each image scene as safe and unsafe according to the workers working environment. The state-of-the-art neural networks successfully perform the binary classification with up to 90% accuracy. Furthermore, feature visualizations, such as Guided Back Propagation, Grad-CAM and different variants of LRP which is successful in showing which pixel in the original image contribute to the diagnosis and to what extent.

Publication Date

4-22-2021

Project Designation

Graduate Research

Primary Advisor

Van Tam Nguyen

Primary Advisor's Department

Computer Science

Keywords

Stander Symposium project, College of Arts and Sciences

United Nations Sustainable Development Goals

Industry, Innovation, and Infrastructure

Explainable Deep Learning for Construction Site Safety

Share

COinS