Talk: Climate Change AI Discussion Seminar: Multimodal AI Approaches for Urban Microclimate Prediction and Building Analysis

This project explores how people perceive building exteriors in urban environments using a combination of traditional survey methods and machine learning. In my presentation, I shared insights from analyzing over 250,000 building images from Singapore, San Francisco, and Amsterdam. By training deep learning models on crowdsourced ratings of 1,200 images across six perceptual attributes — complex, original, ordered, pleasing, boring, and style.

This approach allowed us to uncover spatial patterns in how architectural exteriors are perceived and their connections to building functions, age, and location. Additionally, we applied propensity score matching to examine how buildings influence streetscape perceptions. Our findings highlight that streetscapes with complex, pleasing, and historical building exteriors tend to evoke positive sentiments, while modern and monotonous designs are often associated with feelings of dullness or negativity.

These insights aim to help architects and urban planners better understand public sentiment towards building appearances and their role in shaping urban identity.

@article{liang2024evaluating,
title={Evaluating human perception of building exteriors using street view imagery},
author={Liang, Xiucheng and Chang, Jiat Hwee and Gao, Song and Zhao, Tianhong and Biljecki, Filip},
journal={Building and Environment},
volume={263},
pages={111875},
year={2024},
publisher={Elsevier}
}