InSegt
Code accompanying our paper Content-based Propagation of User Markings for Interactive Segmentation of Patterned Images, published at CVMI workshop at CVPR 2020. An earlier version of the paper is on arxiv: 1809.02226. The CVMI presentation can be seen at the link below.
Our interactive segmentation will, guided by the image content, propagate user markings to similar structures in the rest of the image. This allows easy segmentation of complex structures. For example, see in the animation below, how we obtain the segmentation of an image using just a few markings. The image is a slice from μCT scan of a bee eye. On the left, the input is given as a small subset of manually marked pixels. On the right, the manual labelling is being propagated to the whole image.
In addition we have two examples of fibre detection in a composite material shown below.
