25/09/19 • Kasper Van Lombeek

What we have learned going live with the Zimmo Prijswijzer

Zimmo's Prijswijzer has been running in production for a week. The analysis of the user statistics led us to several interesting insights.

17/09/19 • Kasper Van Lombeek

How does the Zimmo Prijswijzer help to estimate the price of a house?

Zimmo has launched, in collaboration with Rockestate, a new website to help its users to estimate the price of a house. Rockestate has built the model behind the website. This blog post explains the basics of this predictive model.

17/09/19 • Pietjan Vandooren

Open geo-data: Rockestate’s key ingredient

The what, the why and the how with a particular focus on open geo-data

15/07/19 • Kasper Van Lombeek & Pietjan Vandooren

The use and abuse of reconstruction value

Home insurers tend to focus a lot on the reconstruction value of a house to calculate its risks related to a fire insurance. This value suffers from a number of serious flaws, while some more objective building characteristics remain un(der)used.

23/05/19 • Pietjan Vandooren

Belgian real estate market lacks price transparency

Publicly available transaction prices can make the real estate market more transparent. France is showing the way with their recently launched open data portal containg a price history of the past 5 years. Other European countries have similar initiatives. When will Belgium follow?

26/03/19 • Kasper Van Lombeek

Our love-hate relationship with heatmaps and how we use kriging to make them

Although heatmaps come with many pitfalls, we believe it is a good idea to interpolate listed asking prices and turn these interpolations into a heatmap. In this blogpost we go over the concept of a heatmap a bit more in detail, and explain our favourite interpolation technique kriging.

16/01/19 • Mathieu Carette

Fast aerial images part I - tiles

A WMTS (Web Map Tile Service) journey in the belgian coordinate system

07/12/18 • Kasper Van Lombeek

Address matching: as much as possible or better safe than sorry?

Depending on the context of the exercise, different metrics regarding adres matching are important. For some exercises you want to match 100% of the addresses. But in most of the Rockestate exercises, we don't need a 100% matching ratio. We need to be a 100% sure of each match. We illustrate this diffference with some examples.

15/11/18 • Mathieu Carette

Is It Your Solar Panel or Your Neighbour's?

Why 2D needs 3D: correcting building perspective in orthophotos

19/10/18 • Kasper Van Lombeek

Old school feature engineering

Although automated feature engineering is hot today, we still highly value feature engineering based on background knowlegde. We demonstrate this approach with an example.

21/03/18 • Mathieu Carette

The art of computing a building's volume

Challenges encountered while computing volumes and how to cope with them.

05/02/18 • Kasper Van Lombeek & Mathieu Carette

Fosdem 2018 was awesome

Fosdem gets better year after year. It's fun to see lots of known faces at the geospatial devroom, and this time we also gave a talk. We share a link to the video of the talk, and list some talks we liked ourselves.

18/01/18 • Kasper Van Lombeek

Discovering structure in first names by analysing user clicks

Thousands of people clicked on hundreds of names on our recommender system NamesILike.com. The result of applying matrix factorization on these user clicks reveal an unseen structure in our first names.

29/11/17 • Kasper Van Lombeek

Roof solar potential energy

Can we estimate the potential for solar energy for each roof in Flanders based on our 3D models for buildings? And what about the Zonnekaart of the Energie Agentschap Vlaanderen?

10/11/17 • Mathieu Carette

Turning a Jupyter Notebook Into a Presentation

Quickly make an interactive presentation with two lines of CSS

26/10/17 • Mathieu Carette

A Tour of 3D Point Cloud Processing

Combining the strengths of pdal, ipyvolume and jupyter

08/09/17 • Kasper Van Lombeek

How We Accurately Tagged 38.000 Flemish Private Swimming Pools

Using open aerial images, a list of 1000 addresses with swimming pools, and gdal, aws and scikit-learn as data science tools.