Towards the Use of Computer Vision Techniques on Streetscape Imagery to Empower Citizens in the Planning Enforcement Process

Urban streetscapes are often cluttered with intrusive advertising signage, which is typically erected without appropriate planning permission. This paper proposes the deployment of computer vision techniques to automatically identify this type of illicit signage within geotagged and timestamped digital images taken of an urban streetscape from a moving vehicle. Such object detection can underpin a semi-automated workflow for instigating planning enforcement complaints against offending signage at scale. The implemented method utilises deep learning models for object detection on a manually collated and labelled dataset of 1051 images containing illegal advertising signage. The system is evaluated on a batch of acquired streetscape images collected from various urban areas in Dublin, Ireland. The results demonstrate high accuracy and efficiency in detecting non-compliant vinyl banners and property signs. This implementation is offered as an example of the potential for computer vision techniques to mediate new relationships between citizens and local authorities.