IDP at Professorship of Real Estate Development: Building Characteristics Prediction based on Street View Images

Aktuelles

With increasing needs to reduced energy consumption of buildings data collection on large hosuing stocks is important to increase prediction accuracy of energy performance and map less efficient buildings. Reasearchers have developed techniques based on maschine learning to detect and predict building components and charateristics using image data from sources such as Google Street View. Their studies offer promising results for the data collection on large building stocks.

During this project we will use street view images to predict building characteristics such as property use, number of floors, window-to-wall ratio, age. Herefore, we will use well-established maschine learning techniques for image classification, segmentation and object-detection such as multilayer perceptron (MLP) or convolutional neural networks (CNN).

The project can be decomposed into the following tasks:

Study the literature on image classification, segmentation and object-detection to gain an overview over precition methods and valuation techniques. We have listed a couple relevant papers in the real estate field in the reference section at the end!
Download Google Street View Images for given real estate addresses.
Predict building characteristics such as property use, number of floors, window-to-wall ratio, age. You might find more features to predict from the literature.
Validate predictions based on given data.
Visualize prediction accuracy and validation.
Requirements
Independent and thorough operation
Programming experience with Python (e.g., PyTorch, Tensorflow)
Interest / first experience in image classification and segmentation.
Accompanying Lectures
Generally, the accompanying lecture can be selected in consultation with the supervisor from the entire list of courses offered by the Professorship for Real Estate Development. For this IDP, attendance to one of the following lectures is recommended:
Applied Research Methods in Real Estate
Sustainable Real Estate Development
Seminar Real Estate Investment