Deep neural network approach for annual luminance simulations
Yue Liu, Alex Colburn, Mehlika Inanici
Abstract
Annual luminance maps provide meaningful evaluations for occupants’ visual comfort and perception. This paper presents a novel data-driven approach for predicting annual luminance maps from a limited number of point-in-time high-dynamic-range imagery by utilizing a deep neural network. A sensitivity analysis is performed to develop guidelines for determining the minimum and optimum data collection periods for generating accurate maps. The proposed model can faithfully predict high-quality annual panoramic luminance maps from one of the three options within 30 min training time: (i) point-in-time luminance imagery spanning 5% of the year, when evenly distributed during daylight hours, (ii) one-month hourly imagery generated during daylight hours around the equinoxes; or (iii) 9 days of hourly data collected around the spring equinox, summer and winter solstices (2.5% of the year) all suffice to predict the luminance maps for the rest of the year. The DNN predicted high-quality panoramas are validated against Radiance renderings.