Revealing urban secrets with the #MapYourCity Challenge

The #MapYourCity challenge is an initiative supported by ESA Φ-lab in the framework of the AI4EO challenges. The 2024 edition took place from 02 April to 14 July, bringing together researchers and coders, and driving a positive change through the use of artificial intelligence technology (AI) and Earth observation (EO) data for automated building age detection. 

Every building has a story. From its requirements to characteristics such as architectural style, construction techniques and design philosophies, knowing the condition of a building is essential to maintain its structural integrity and safety. In particular, the age of a building is a very important variable to consider during renovations or preservation efforts. Age-related structural and safety issues may require a rapid and tailored action that prevents potential hazards and improves urban city planning policies.

As cities continue to grow, comprehensive and organised monitoring of building age becomes a very difficult task. Manual sampling and strew-view observations are often tedious and time-consuming, being compromised by ongoing construction and demolition projects that alter the urban landscape and obstruct clear views of buildings.

In that sense, the #MapYourCity Challenge emerged as a way to revolutionise the monitoring of urban environments. Supported by ESA Φ-lab, together with Novaspace, EarthPulse, Sinergise and Planetek Italia, this AI4EO challenge took place from 2 April to 14 July 2024, leveraging the use of EO data and AI automation and offering a detailed and diverse perspective on our cities, from street level to satellite view.

Participants were challenged to create their own innovative solution, by training a deep-learning model capable of accurately estimating the construction year of any given building. To achieve this, they were provided with a training dataset – a large group of data used to train AI models, so they can process information and accurately predict outcomes.

This particular dataset, curated by MindEarth, included information from urban buildings in five different countries and over a 100-year timespan, such as building footprints (from EUBUCCO), date of building construction, street-level imagery of the building façade (provided by Mapillary), medium-resolution cloud-free Sentinel-2 images, and very-high resolution (VHR) images by ESA Third Party Mission Airbus Pléiades. Ultimately, the goal was to estimate the age of a building using only top-view perspectives, so that the developed system could be applied at scale.

A total of 123 teams registered for the challenge, with 30 teams actively participating and more than 300 submissions. The winners of the challenge were announced during URBIS2024: the third place was given to Caroline Arnold, from DKRZ (Germany); Tran Hoang Ba from Axelspace (Japan) won the second prize; and the grand prize was awarded to Eric Park, Hagai Raja Snulingga and Steve Immanuel from TelePIX (South Korea). The winners were rewarded with a cumulative prize of € 5000.

Estimating the age of a building is, without a doubt, a task improved by the outcome of this challenge. Nevertheless, the type of approach developed during the competition can also be applied to other important characteristics in a building: the BEE-AI project, funded by ESA and developed by MindEarth, “aims to enrich existing energy certification processes by offering a comprehensive view of urban energy efficiency at the level of individual buildings.”

Nicolas Longépé, Earth Observation Data Scientist at Φ-lab, comments the outcome of the initiative: “By bringing together the worlds of artificial intelligence and Earth observation, this challenge promoted not only the growth and engagement of the AI4EO community, but also provided a platform for researchers and developers to showcase their work and make a tangible impact in solving the challenges posed by increased urban growth and the lack of appropriate methods to monitor the conditions of buildings in an optimal way.”

Know more about #MapYourCity and other Φ-lab-supported challenges at www.ai4eo.eu.

To know more: ESA Φ-lab, MindEarth, AI4EO Challenges

Photo courtesy of Unsplash/Mohit Kumar