Scholarlink

<p>Rapid urban expansion and rising housing prices have created significant social and economic challenges in many cities, exposing the limitations of traditional, resource-intensive data collection methods for urban planning. Remote sensing technologies, such as satellite imagery, offer a cost-effective alternative for data collection. Incorporating street view imagery can enhance the quality of collected information by providing a human perspective on the urban environment, often inaccessible through standard remote sensing methods. This research aimed to develop a deep learning classification model using a multi-modal fusion of satellite and street view imagery to predict urban metrics of housing price, housing density, and green area coverage. Focusing on Lisbon, a European city facing rising housing prices, this research used EfficientNetB0, a pre-trained model originally used for object recognition on the ImageNet dataset, which successfully generalized its learning to interpret urban imagery. The findings highlight the potential of integrating pre-trained models within deep learning frameworks for urban analysis. This approach leverages low-cost, readily available data, providing a scalable alternative to traditional methods and a foundation for developing predictive tools for urban metrics.</p>

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

Predicting housing price, housing density, and green area coverage from combined satellite and street view imagery using deep learning

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月11日

查看全文
  • Ana Beatriz de Figueiredo Oliveira,
  • Mauro Castelli,
  • Esra Suel

摘要

Rapid urban expansion and rising housing prices have created significant social and economic challenges in many cities, exposing the limitations of traditional, resource-intensive data collection methods for urban planning. Remote sensing technologies, such as satellite imagery, offer a cost-effective alternative for data collection. Incorporating street view imagery can enhance the quality of collected information by providing a human perspective on the urban environment, often inaccessible through standard remote sensing methods. This research aimed to develop a deep learning classification model using a multi-modal fusion of satellite and street view imagery to predict urban metrics of housing price, housing density, and green area coverage. Focusing on Lisbon, a European city facing rising housing prices, this research used EfficientNetB0, a pre-trained model originally used for object recognition on the ImageNet dataset, which successfully generalized its learning to interpret urban imagery. The findings highlight the potential of integrating pre-trained models within deep learning frameworks for urban analysis. This approach leverages low-cost, readily available data, providing a scalable alternative to traditional methods and a foundation for developing predictive tools for urban metrics.