Geometry in Active Learning for Binary and Multi-class Image Segmentation

We propose an Active Learning approach to training a segmentation classifier that exploits geometric priors to streamline the annotation process in both background-foreground and multi-class segmentation tasks that work in 2D and 3D image volumes. To this end, we use smoothness priors not only to select voxels most in need of annotation but to guarantee that they lie on 2D planar patch, which makes it much easier to annotate than if they were randomly distributed in the volume. We evaluate our approach on Electron Microscopy and Magnetic Resonance image volumes, as well as on natural images of horses and faces. We demonstrate a marked performance increase over state-of-the-art approaches.