Figure 2.

MitoSegNet Visually Outperforms Feature-Based Segmentation Approaches

On the left side, four images of elongated, fragmented, tubular, and mixed C. elegans mitochondria and their respective ground truth are shown. The masks on the right show the results of MitoSegNet and the four segmentation methods applied to each image, displaying the false negative segmentation in red, the false positive segmentation in turquoise, the true negative segmentation as black, and the true positive segmentation as white. The yellow arrows indicate areas in which false segmentation occured. The scale bar is 5 μm.