A basic analysis on urban landscape continuity in a lowland urban heritage using deep learning-based method
Architectural intervention in urban heritage area is subject to numerous parameters making it a time-consuming process. Urban façade analyses are also one of the required long-term tasks held by the architect, especially in urban heritage area where pressure concerning the neighborhood harmony is often faced.
To address this issue, we here present a computer vision method for an automatic evaluation of the urban façade, by comparing a set of façade’s pictures. Our target area is Hizenhamashuku, in a “preservation area of traditional buildings” located in Kashima city, which is a typical lowland city in Saga prefecture.
In this project we explore possibilities to boost the performance of urban facades study using a deep learning method. We developed an algorithm able analyze pictures of buildings from different historic eras with different historic styles, regarding any selected feature.
First, an objective feature, such us the orientation of the building which, having a unique parameter, prevent from bias and thus its results can be used as reference. Next in order, a more subjective parameter such as the quality of insertion is tested, results are quantified and compared in order to evaluate the algorithm performance and enhance it in further research