9 August 2022, Geneva, Switzerland – In July 2019, the United Nations Institute for Training and Research (UNITAR) signed a Memorandum of Understanding with Wuhan University (WHU) to leverage Artificial Intelligence (AI) research and innovation capacities for remote sensing through the United Nations Satellite Centre (UNOSAT). Two years and a pandemic later, the collaboration keeps strong, with new WHU students starting their traineeships at the UNOSAT headquarters in Geneva.

“This collaboration with Wuhan University is a great opportunity for us to increase our research capacity to investigate the potential of Artificial Intelligence and Machine Learning for geospatial analysis. We are very happy to welcome the students, give them access to the UNOSAT facilities and leverage their expertise.” Einar Bjorgo, Director of UNOSAT

WHU is a leading university in Asia, with over 7600 professors and staff and 34 schools. It is well known for remote sensing, a subject taught for over 30 years and for which the university ranked first in the Shanghai Ranking’s Global Ranking of Academic Subjects in 2018.

UNOSAT and WHU share the common goal to develop capacity in training and humanitarian support using Geospatial Information Technologies (GIT). This is accomplished through the cooperation on research and innovation in the use of GIT in support of the 2030 Agenda for Sustainable Development, providing training to developing countries in the use of GIT, as well as awareness raising activities on the benefits of GIT.

Hongruixan Chen started his traineeship with UNOSAT in May 2021 while finishing his master’s degree in Photogrammetry and Remote Sensing at WHU. Although he was working remotely, Chen had full access to the data and facilities hosted in Switzerland.

Collaborating closely with the rest of the UNOSAT team, Chen’s research focused on using innovative deep learning, specifically tasking large-scale satellite images for building damage assessment caused by natural hazards. This experience allowed him to deepen his knowledge on this topic and also learn to deal with large quantities of large-scale satellite images since deep learning requires substantial amounts of data.

“My previous work, just focus on the general changes, such as the land cover and land use changes. Here building damages are a very specific type of change: after a disaster occurs, the building is destroyed, or it's just minor damaged. And secondly I learned how to deal with very large-scale data. In my previous research work, I was only involved in handling small-scale data. But during these traineeship I had to handle tens of thousands of remote sensing images.” Hongruixuan Chen

The work he accomplished during his traineeship resulted in a better understanding of using a transformer-based architecture, a recent advancement in AI research, for remote sensing interpretation. The results were published in a joint publication by UNOSAT, WHU, and CERN and then presented by Chen in July 2022 during the International Geoscience and Remote Sensing Symposium with a presentation titled “Dual-tasks Siamese Transformer Framework for Building Damage Assessment”.

“My traineeship experience was perfect for me. I truly applied remote sensing technology into practical applications, contributing to humanitarian assistance. Besides, the team and the facilities are great. The research field in my PhD degree will also be the damage assessment, so this is particularly important preliminary work.” Hongruixuan Chen

Through this framework, UNOSAT welcomed 9 students. So far, the students have helped advance research on AI for remote sensing on various topics, including shelter mapping, building footprint, and flood detection. They also supported UNOSAT throughout the entire AI live cycle starting from applied research to deployment, experimenting through data preparation, modelling, development of validation protocols, etc.

Leveraging the collaboration with WHU and other partners is a pillar of the advancement of AI research for geospatial analysis and remote sensing.  The increased capacity allows UNOSAT to pioneer in this field and apply the conclusive results directly in the humanitarian support operations. The deployment of the UNOSAT FloodAI model and its integration into the Emergency Mapping service operations was the first example of how AI can concretely benefit disaster response by working at scale rather than at capacity.

To learn more about the outcomes of this collaboration, follow UNOSAT on Twitter and visit UNOSAT’s website.