THESIS
Special thanks to David Andrés León, Justyna Szychowska, Dai Kandil, Sara Kessba, and Lucas Sentís Fuster.
Thank you to Prof. Wassim Jabi and Eduardo Rico for their invaluable feedback during our final jury on Sep 18th, 2023.
This thesis focuses on using graph machine learning (GML) for node and edge classification on the web. Users outline their building geometry using the leaflet.js map below that is then processed using Rhino.Compute and a trained graphSAGE model. The result is returned and displayed in Three.js. The example below uses this methodology for PREDICTING EGRESS FOR MULTI-STORY RESIDENTIAL BUILDING
We have applied knowledge gained from two IAAC MaCAD 2022/23 courses.
CLOUD-BASED DATA MANAGEMENT emphasized front-end development using HTML, CSS, and JavaScript to interface with Rhino.Compute to control Rhino3D or Grasshopper3D geometry within a web browser.
GRAPH-MACHINE LEARNING utilized graphs composed of edges and nodes to represent buildings. Machine learning techniques were employed to train a model for predicting node and edge classification.
This button sends a JSON containing your inputs to the backend. JSON is a lightweight data storage interchange format that uses key-value pairs.
It will return another JSON containing all of the nodes and edges, each classified as Units, Egress, or Corridor. A graph is displayed below built in three.js using this data.
Load times may take up to 1 minute. Longer requests are currently not allowed.
Please wait two minutes for the server to wake up after your first request
