Deep Learning Approach for Analyzing Urban Construction Material Stocks and Evaluating Environmental and Economic Benefits of Alternative Building Material

Abstract

Urban centers contribute substantially to global greenhouse gas emissions, and with ongoing urbanization, the demand for infrastructure construction is set to rise. The construction and building sector significantly impact the environment, responsible for 36% of the world's energy consumption and 39% of its CO2 emissions. Annually, the global production of concrete and cement reaches 17.7 Gt and 4.1 Gt, respectively, making cement the second most utilized substance after water. Cement production alone contributes to 9-10% of worldwide CO2 emissions related to energy.This thesis comprises of two research projects to mitigate the environmental footprint of urban development. First, we develop the computer vision based deep learning model to estimate building material stocks (MS) in cities. Traditional methods of estimating MS often falter due to the lack of granular building data. We propose a novel solution by employing deep learning to derive MS estimates from readily available aerial and street-view imagery. Our methodology involves the development of two deep learning models that adeptly classify building types and predict floor areas, respectively. The model demonstrates reliable performance in predicting MS. Second, this thesis explores the environmental and cost benefits of alternative building materials: mass-timber and low-carbon cement. We create future scenarios and compare mass-timber and low-carbon cement in terms of carbon mitigation and cost savings. Results indicate that mass-timber buildings offer net-negative emissions and cost savings compared to conventional concrete-frame buildings.