||In many biomedical applications, an in-sufficient amount of annotated images restricts successful training of Convolutional Neural Networks (CNNs). Although transfer learning and various image augmentation techniques can often help training CNNs from a relatively small image set, in many applications, the objects of interest cannot be accurately delineated due to their fuzzy shape, image quality or a restriction in time, knowledge or experience of the expert performing the annotation. I will present an approach for training segmentation CNNs from images with minimal annotation. This limited annotation consists merely of centerlines or center points of objects of interest. I will demonstrate this approach in the application of virus segmentation in a challenging transmission electron microscopy image dataset.