The aim of the course is to unlock and channel the creative potential of architects in the era of AI. This will be done mainly by providing valuable resources and methods for quantitatively curating and evaluating visual architectural data. This course is delivered at the Faculty of Architecture and the Built Environment (Delft University of Technology) as a 5EC course in the third quarter of the first year of the Architecture Master's track: Master 2, Q3. Flyer.
- Instructor: Seyran Khademi.
- Assistants: Casper van Engelenburg, Fatemeh Mostafavi, Pablo G. Morato, Julien Vuillamy
| Tutorial | Learning objectives |
|---|---|
| T0_Intro_to_Python_and_Colab | Get familiar Python programming in Google Colab |
| T1_From_Code_to_Canvas | Use Google Collab and run code Create and print most common data types Create and manipulate polygonal shapes Plot polygonal shapes Use for loops and functions |
| T2_From_Numbers_to_Plots | Make use of CSV files to create a DataFrame Cleaning Data (reading, sorting, and selecting) Plotting FloorPlans |
| T3_From_Geometries_to_Graphs | Define a graph Create, manipulate, and visualize a graph in Python Describe the access graph of a floor plan Extract (apartment-level) access graphs from the IFC building elements. |
| T4_From_Footprints_to_Photos | Visualize and interpret building+context representations Automatically collect aerial images and create a customized dataset Locate building footprints from geographical information |
| T5_From_Photos_to_Embeddings | Generate image embeddings from pre-trained foundation models Compute the cosine similarity between embeddings Interpret building+context representations |
| T6_From_Images_to_3D_Understanding | Understand opportunities of AI in computer vision and photogrammetry Introduce the concept of data fusion when working with multiple modalities |
| T7_From_Graphs_to_Similarity | Investigate floor layout similarity using pre-trained deep neural networks specialized in extracting layout-specific features |
| T8_Similarity_Urban_Scale | Visualize and interpret urban representations at scale Create a customized dataset of aerial and street view images Generate an urban similarity score by combining similarity scores computed from both aerial and street view images |
| W0_WELL_for_residential | Create a visual narrative to verify some health and well-being concepts of WELL for residential standard |