Fig. 4
From: Deep learning-based analysis of macaque corneal sub-basal nerve fibers in confocal microscopy images
The best performing U-Net architecture used in these analyses, titled deepNerve. Depiction of the U-Net architecture showing both the encoding and decoding paths and their dimensions. For the given input image (left), it outputs two images giving the probability score, at each pixel, of belonging to one of the two classes (right). The final softmax layer ensures these are normalized and can thus be interpreted as probabilities