Our goal is to analyze how rivers have developed over time using historical aerial images. However, historical aerial images are greyscale and captured over almost a hundred years. In the same period, camera technology has advanced a lot, which increases the difficulty of the task. Therefore, we seek to improve semantic segmentation of such images.
Our team consists of Saeid Shamsaliei from the Norwegian University of Science and Technology (NTNU), Jo Halvard Halleraker from NTNU and the Norwegian Environmental Agency, Professor Knut Alfredsen from NTNU, Anders Foldvik from the Norwegian Institute for Nature Research and Odd Erik Gundersen from NTNU and Aneo AI Research.
The benchmark code can be found on the project's Github page.
The dataset and Code Are released under Creative Commons CC-BY-SA 4.0 LICENSE.