About CRIM

Évia - CRIM’s expertise tackles a real estate property assessment software

JLR, a company specialized in land title solutions, approached CRIM’s research team with the objective of optimizing their real estate property assessment software, Évia. This platform enables real estate agents or individuals to obtain a personalized report outlining the main features and market value of any property in Quebec, automatically and in a few minutes.

How it works

At first, this software was designed following a statistical model that analyzes and combines data in order to come up with a market price bracket and a median value of similar properties recently sold in the same area. JLR now wishes to optimize this model in order to obtain more precise results. The company has asked two of CRIM’s teams to work on this project: Vision and Imaging and Emerging Technologies and Data Science. Both teams will collaborate to build a new structure for the software that will be based on machine learning and neural networks.

Vision and Imaging

JLR already possesses a bank of over 1.2 million photos of property facades in its database, and has started a complete visual capture of the province through its Jakarto project. The Vision and Imaging team will produce an algorithm that can determine automatically, from photographs, various physical characteristics from properties: type of exterior surfacing, size of windows, presence of a garage, etc. This information will serve to fill parts of Évia’s report automatically.

Emerging Technologies and Data Science

The Emerging Technologies and Data Science team will integrate the visual data and will combine it with existing data on each property taken from the JLR database, real estate transaction information has been collected for over 30 years. The team will use these datasets to construct a more precise market value prediction model. The model will be conceived such as to be able to calculate a market value bracket even if some of the data on the property is missing. It will also be able to signal property transactions that stand out from the average trend.

Final goal

In order to respond to JLR’s needs, CRIM seeks to replace the software’s current statistical approaches with machine learning processes in order to obtain more precise results about the value of each property, while augmenting the system’s robustness. CRIM will present a functional prototype in the fall 2018 that JLR will then be able to integrate into its website. 


Upcoming event

  • Séminaire R-D : L'art de la cyberrésilience: les leçons du secteur bancaire
    30 January 2020 11:00
    CRIM (405, avenue Ogilvy, bureau 101, Montréal)
    Une conférence du Prof. Benoît Dupont, Titulaire de la Chaire de recherche du Canada en cybersécurité,Titulaire de la Chaire de recherche en prévention de la cybercriminalité, Directeur scientifique – Réseau intégré sur la cybersécurité (SERENE-RISC), Université de Montréal.
  • #CRIMtravels ????Joao Monteiro, automatic speech processing intern supervised by Jahangir Alam, presents an article at… https://t.co/7bsFunP7na
  • Antoine Normand, PDG de BlueBear a discuté d'un sujet sensible lors de la Journée Techno sur la Désinformation à l’… https://t.co/FlBn1vdEaE

Recent Publications

  • Deep learning speech recognition for low-resource languages

  • Forage de données géospatiales, quelques applications