HUN-REN SZTAKI success at IEEE GRSS 2026 Data Fusion Contest

Jamil Ghazal, a first-year PhD student at the Artificial Intelligence Research Laboratory of HUN-REN SZTAKI, has won the IEEE GRSS 2026 Data Fusion Contest, an international scientific competition. The award-winning work was carried out in collaboration with ELTE; the team members were Ghazal Jamil, András Jung, and Vera Könyves. András Jung and Vera Könyves are Jamil’s PhD supervisors. The IEEE Geoscience and Remote Sensing Society (IEEE GRSS) is one of the leading international professional communities in Earth observation, remote sensing, and radar satellite image processing.

The Data Fusion Contest, announced annually by the organization, is a well-known scientific challenge in the field, where participants develop artificial intelligence and data analysis-based solutions for interpreting Earth observation data. The 2026 competition was jointly organized by IEEE GRSS and Capella Space. Capella provided high-resolution radar satellite data to the participants. A distinctive feature of this year’s competition was that there was no narrowly predefined task. Instead, participants had to identify a scientifically relevant problem themselves from the available radar time-series data and develop their own method to address it. The award-winning solution of the HUN-REN SZTAKI–ELTE team is a self-supervised artificial intelligence method called TRISAR.

TRISAR is a self-supervised AI framework for interpreting temporal SAR image series. Instead of relying on manually annotated training data, the method learns from repeated observations of the same geographic area and builds a feature representation in which normal temporal variability stays compact, while suspicious deviations stand out. The approach combines triplet-based deep metric learning, SAR-specific data augmentation, and a ConvNeXt-based neural backbone, then uses the learned representation for three practical tasks: ranking suspicious image pairs, analyzing how a region evolves over time, and localizing the most relevant anomalous areas in the images.

Jamil will present the team's work at the dedicated Data Fusion Contest session of IGARSS 2026 in Washington, D.C. In addition, the team will publish their paper in the IGARSS 2026 Proceedings, receive IEEE Certificates of Recognition, and co-author an open-access IEEE JSTARS journal article on the contest results. This makes the recognition especially significant, as it comes from one of the most highly regarded international communities in remote sensing, whose flagship event, IGARSS, is the field's leading annual conference.