Semantic segmentation is the task of assigning a class to each pixel in an image, which is an important objective in a number of domains. For instance, doctors might want to subdivide medical imagery into different portions in order to study the structure of bones and organs. Furthermore, autonomous vehicles utilize cameras to observe their surroundings. The images produced by the cameras are segmented and the vehicles navigate using these segmented images.
The UiT Machine Learning Group has focused on advancing semantic segmentation in remote sensing and in the medical domain, while simultaneously producing innovative semantic segmentation methodology. In particular, we have moved uncertainty modelling for deep learning based models forwards in both remote sensing and in the medical domain. Also, we have proposed innovative methods for producing transparent and interpretable deep learning based models.
- Michael Kampffmeyer, Arnt-Børre Salberg and Robert Jenssen
- Kristoffer Knutsen Wickstrøm, Michael Kampffmeyer and Robert Jenssen