Mapping the extent of giant Antarctic icebergs with deep learning

13 November 2023

In this study, the authors propose a deep neural network to map the extent of giant Antarctic icebergs in Sentinel-1 images automatically. While each manual delineation requires several minutes, their U-net takes less than 0.01 s. In terms of accuracy, they find that U-net outperforms two standard segmentation techniques (Otsu, k-means) in most metrics and is more robust to challenging scenes with sea ice, coast and other icebergs. The absolute median deviation in iceberg area across 191 images is 4.1 %.

The press release by the University of Leeds can be found at: https://www.leeds.ac.uk/news-technology/news/article/5448/ai-can-map-the-outline-and-area-of-giant-icebergs

Contact: Anne Braakmann-Folgmann (anne.braakmann@uit.no)