Metaverse Apartment
Retrieval Challenge


In contemporary society, the search for new accommodation has become a frequent and sometimes necessary undertaking. The reasons behind this movement are diverse, ranging from job relocations and educational pursuits to lifestyle changes and personal preferences. The traditional approach to finding a suitable apartment involves a time-consuming process, often necessitating physical visits to assess the accuracy of online advertisements.

However, with the advent of the Metaverse, exploring apartments virtually emerges as a promising and efficient alternative.

Visiting apartments in the Metaverse offers a compelling green alternative to the traditional approach. By leveraging virtual reality technologies, individuals can explore potential living spaces from the comfort of their current location, eliminating the need for extensive and expensive travel. This not only saves time and resources but also reduces the carbon footprint associated with commuting. The metaverse thus presents an innovative solution to streamline the apartment search process, aligning with the growing global emphasis on sustainable practices.

Recognizing the limitations of existing web-based recommendation platforms for this purpose, a new problem is identified: text-to-apartment recommendation. This entails developing systems capable of ranking apartments based on their relevance to a user’s textual query expressing specific interests.

To raise awareness on this problem, we are currently organizing a challenge, within the CV4Metaverse workshop co-located with ECCV 2024, the 18th European Conference on Computer Vision.


The goal is to rank Metaverse apartments (furnished 3D indoor scenes) based on their relevance to a textual query (formal descriptions, and realistic user queries).

Join the Google Group!

Join the Challenge!

Dataset and Resources

The dataset contains around 6800 3D apartments, each annotated with a description covering the amount of rooms, their category, and all the present furnishing with ample details.

Around 6000 of those apartments and their descriptions are part of the train/validation/public test split (available on GitHub).

The additional 800 apartments will represent the private test set.

Moreover, we will extend the evaluation both with respect to the descriptions and to more informal user queries we are collecting.

Data and Baseline

Important Dates

Tentative timeline (may be subject to changes).

25 March 2024

The challenge website is up and the challenge starts (only 6000 apartments are public).

01 July 2024

Evaluation on a private held-out set.

30 July 2024

Participants submit a final report describing their system.

29-30 September 2024

The ranking on the private test set will be made public, and winners will be announced at a dedicated event.
Selected participants will hold a brief presentation describing their methodology.


Ali Abdari

PhD Student
University of Udine, Italy
University of Naples Federico II, Italy

Alex Falcon

Postdoc Research Fellow
University of Udine, Italy

Beatrice Portelli

PhD Student
University of Udine, Italy
University of Naples Federico II, Italy

Giuseppe Serra

Associate Professor
University of Udine, Italy

Richard Zhang

Distinguished Professor
Simon Fraser University, Canada

Barbara Roessle

PhD Student
Technical University of Munich, Germany

Maria Pegia

PhD Student
Reykjavik University, Iceland
CERTH, Greece


To join the challenge, fill-in the following google form:

Join the Challenge!

To keep updated, join our google group:

Join the Google Group!

For any other information, send us an email at: