ThermoVision is a predictive tool that leverages machine learning to gauge thermal comfort in urban landscapes, allowing us to understand the correlation between our built environment and the thermal comfort.
There is inconsistent outdoor thermal comfort due to varied building topologies, building materials and vegetation. Understanding the thermal landscapes is challenging due to their complexity. The objective of the project is to explore and quantify the relationship between immediate environmental conditions and visually perceived outdoor thermal comfort. ThermoVision is a machine learning tool to predict and assess the thermal comfort experienced at specified outdoor settings, facilitating a better understanding of outdoor comfort dynamics in urban areas.
360 degree street level imagery were extracted and semantically segmented using SegFormer, and features of the built environment were correlated and trained with a Random Forest Regression Model which was tested to predict thermal comfort in different locations.