Location is key – especially when it comes to real estate value. “ImmBild” aims at developing a new method for estimating property value using computer vision of satellite data.
External Staff
David Koch,University of Applied Sciences Kufstein
Mario Döller, University of Applied Sciences Kufstein
  • University of Applied Sciences Kufstein
Austrian Research Promotion Agency
01/01/2017 – 12/31/2020
Logo ImmBild

Real estate value and image analysis

The value of any real estate property depends most likely on its location. „ImmBild“ is an innovative approach, combining real estate industry specific knowledge and automated image analysis to find new ways to determine rent or real estate prices. An automated image analysis of satellite pictures in combination with geographical coordinates helps determine the value and quality of real estate location – data that up until now has not been taken into account. 
Our newly suggested approach wants to increase transparency on the real estate market for all market participants and facilitate property evaluation and rent regulation.

A new method for real estate appraisal

Traditionally real estate appraisal is based on hedonic regression, meaning an analysis taking into account certain characteristics of location such as nearby schools, shopping opportunities or socio-demographic data. Our new method, still going further, also includes regional characteristics for the classification and evaluation of location such as neighborhood, green spaces, infrastructure, recreation areas or land to building ratio. All of the above are identified by our data analysis tools and can be taken into account for real estate appraisal.

Automated image analysis of satellite imagery

Since we have big amounts of satellite pictures at our disposal, those are highly qualified for machine learning. Our project uses neural networks or Deep Learning to identify and classify different elements such as streets, water, trees or houses. Built on those segmentations and qualifications, combined with traditional regression methods, long-term criteria to determine and evaluate location can be established.


Despotovic, Miroslav, Sakeena, Muntaha, Koch, David, Döller, mario, & Zeppelzauer, Matthias. (2017). Predicting Heating Energy Demand by Computer Vision. Computer Science - Research and Development, 33, 231–232.

Press articles

Studienprojekt: Hausbesichtigung per Satellit

Publication date: 11.10.2017
Medium: Die Presse

You want to know more? Feel free to ask!

FH-Prof. Dipl.-Ing. Mag. Dr. Matthias Zeppelzauer
Senior Researcher
Media Computing Research Group
Institute of Creative\Media/Technologies
Department of Media and Digital Technologies
Location: Heinrich Schneidmadl-Straße 15
P: +43/2742/313 228 652