Understanding U-Net

U-Net is developed for the task of semantic segmentation. When a neural network is fed images as inputs, we can choose to classify objects either generally or by instances. We can predict what object is in the image (image classification), where all objects are located(image localization/semantic segmentation), or where individual objects are located (object detection/instance segmentation). The figure below shows differences between these computer vision tasks. To simplify the matter, we only consider classification for only one class and one label (binary classification). <a href="https://towardsdatascience.com/understanding-u-net-61276b10f360">Learn More</a>