[1Win4] Poster session 1

[1Win4-89] Designing Indicators for AI-Based Streetscape Evaluation Systems

〇Takayoshi Yoshimura¹, Keiko Uemura¹, Keisuke Otaki¹ (1.Toyota Central R&D Labs., Inc.)

Keywords:Landscape evaluation, Dataset preparation

The streets where people come and go are essential infrastructure that supports a city, and the quality of their design significantly impacts various aspects such as the economy, transportation, public safety, and disaster prevention. Therefore, several methods have been proposed to evaluate the quality of streets and compare them with those in other cities. However, these evaluations typically require on-site surveys and interviews, making it extremely labor-intensive to assess all the streets in a city. To address this challenge, we have designed evaluation indicators based on street images, with the aim of enabling automatic assessment through AI technology using street viewing services such as Google Street View and geotagged street images posted on social media. The requirements that human-centered streets should fulfill were organized based on architect Jan Gehl's book Cities for People, and a requirement list titled "Qualities of Street Space for People (QSSP)" was developed. We present the results of manually calculating the QSSP scores using street images from the Asakusa area as a case study. Furthermore, we discuss a comparison of our method with existing automatic street evaluation approaches.