Scholarlink

Most companies and organisations face notorious challenges in space management. The Internet-of-Things (IoT) networks attempt to pave the way to get space allocation information about the environment using sensors. However, the deployment of sensors and IoT tags is not cost-effective for large offices. Therefore, this work aims to develop a novel privacy-preserving framework based on thermal imaging cameras and machine learning (ML) techniques in order to monitor space allocation remotely. Hence, to utilise and manage dedicated and unused space appropriately. The work analyses the progress of the proposed methodology via several Deep Learning (DL) techniques based on convolutional neural networks (CNN) and vision transformers (ViT). The experimental results indicated that Visual Geometry Group (VGG) 16 model outperforms other models such as ViT, ResNet50, AlexNet and etc. In addition, a web application has been developed to select a model that is preferred to identify human occupancy in real-time and, therefore, to process space monitoring and utilisation remotely in large office settings.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Novel Thermal Imaging and Machine Learning Based Privacy Preserving Framework for Efficient Space Allocation, Utilisation and Management

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Maria Bruevich,
  • Nilupulee A. Gunathilake,
  • Mandar Gogate,
  • Adeel Hussain,
  • Bin Luo,
  • Jinchang Ren,
  • Amir Hussain,
  • Fengling Jiang,
  • Kia Dashtipour

摘要

Most companies and organisations face notorious challenges in space management. The Internet-of-Things (IoT) networks attempt to pave the way to get space allocation information about the environment using sensors. However, the deployment of sensors and IoT tags is not cost-effective for large offices. Therefore, this work aims to develop a novel privacy-preserving framework based on thermal imaging cameras and machine learning (ML) techniques in order to monitor space allocation remotely. Hence, to utilise and manage dedicated and unused space appropriately. The work analyses the progress of the proposed methodology via several Deep Learning (DL) techniques based on convolutional neural networks (CNN) and vision transformers (ViT). The experimental results indicated that Visual Geometry Group (VGG) 16 model outperforms other models such as ViT, ResNet50, AlexNet and etc. In addition, a web application has been developed to select a model that is preferred to identify human occupancy in real-time and, therefore, to process space monitoring and utilisation remotely in large office settings.