Machine Learning for Daylighting: Predicting Long-term Luminance Maps Using Deep Neural Networks – IBPSA-USA Research Committee

Date of Recording:
Long Description

Annual luminance maps provide meaningful evaluations for occupants’ visual comfort, preferences, and perception. However, acquiring annual luminance maps require labor-intensive and time-consuming simulations or impracticable long-term field measurements. In this Research Committee talk, we will present a novel data-driven machine learning approach that makes annual luminance-based evaluations more efficient and accessible. The methodology is based on predicting the annual panoramic luminance maps from a limited number of point-in-time high dynamic range imagery by utilizing a deep neural network (DNN). Unlike the fixed camera viewpoint of perspective or fisheye projections that are commonly used in daylighting evaluations, panoramas (with 360° horizontal and 180° vertical field of view)  allow full degree-of-freedom in camera roll, pitch, and yaw, thus providing a robust source of information for an occupant’s visual experience in a given environment. The DNN predicted high-quality panoramas are validated against Radiance (RPICT) renderings using a series of quantitative and qualitative metrics. The most efficient predictions are achieved with 9 days of hourly data collected around the spring equinox, summer and winter solstices (2.5% of the year) to predict the luminance maps for the rest of the year. The results clearly show that practitioners and researchers can efficiently incorporate long-term luminance-based metrics over multiple view directions into the design and research processes using the proposed DNN workflow. We share a public dataset of annual HDR panoramic luminance maps and the machine learning codebase to enable reproducibility and future explorations (https://github.com/yueAUW/neural-daylighting.git).

Learning Objectives​:

  • Understand the basic steps of developing a deep learning framework to conduct architectural lighting predictions.
  • Understand how to generate annual luminance maps from a small subset of data using deep neural networks.
  • Learn the impact of the weather and seasonal changes on the accuracy of the prediction outcome, given a fixed data collection period.
  • Learn the potential of machine learning applications in architectural lighting studies.

Speaker Bio:

Dr. Yue Liu is a building technology researcher and a software developer. Her research lies at the intersection of artificial intelligence, physically based simulations, computer graphics, and architectural design/lighting.  She is currently a research scientist at Wormpex AI Research where she builds the state-of-the-art solutions and develops tools to simulate large-scale retail environment in order to improve customers shopping experiences (thermal/visual comfort) and reduce operation costs (energy efficient). Yue has a Ph.D degree in Computation Lighting from University of Washington where she develops end-to-end machine learning frameworks for rapid modeling of daylit environments. Prior to that, she holds a Master of Building Science degree from University of Southern California, and a Bachelor of Architecture degree from Hunan University in China. She dedicates her career to advance the field of building technology and research by developing advanced, yet accessible tools for all users.

machine learning, ML, daylighting, annual luminance maps, deep neural networks, deep learning framework, architectural lighting