The U-Net is a convolutional neural network designed for biomedical image segmentation. It consists of a contracting path for context capture and an expanding path for precise localization. The network is trained end-to-end with data augmentation, allowing it to perform well with few annotated images. It outperforms previous methods in segmentation tasks, such as neuronal structures in electron microscopy and cell tracking in light microscopy. The network is efficient, with segmentation of a 512x512 image taking less than a second on a GPU. It uses elastic deformations for data augmentation to learn invariance to variations in tissue. The architecture includes a symmetric expanding path, allowing seamless segmentation of large images through overlap-tile strategies. The network also uses a weighted loss function to handle touching cells. It was successful in the ISBI challenges for both EM segmentation and cell tracking. The network is implemented in Caffe and available online. The U-Net is effective for various biomedical segmentation tasks and can be applied to many more tasks.The U-Net is a convolutional neural network designed for biomedical image segmentation. It consists of a contracting path for context capture and an expanding path for precise localization. The network is trained end-to-end with data augmentation, allowing it to perform well with few annotated images. It outperforms previous methods in segmentation tasks, such as neuronal structures in electron microscopy and cell tracking in light microscopy. The network is efficient, with segmentation of a 512x512 image taking less than a second on a GPU. It uses elastic deformations for data augmentation to learn invariance to variations in tissue. The architecture includes a symmetric expanding path, allowing seamless segmentation of large images through overlap-tile strategies. The network also uses a weighted loss function to handle touching cells. It was successful in the ISBI challenges for both EM segmentation and cell tracking. The network is implemented in Caffe and available online. The U-Net is effective for various biomedical segmentation tasks and can be applied to many more tasks.