deep-learningCNNsegmentationmedicalessentials

U-Net Convolutional Networks for Biomedical Image Segmentation

Reviewed on Feb 27, 2017 • https://arxiv.org/pdf/1505.04597.pdf

Reference : O. Ronneberger, P. Fischer, T. Brox. U-Net: Convolutional Networks for Biomedical Image Segmentation. Proceedings of MICCAI 2015, p.234-241.

Famous 2D image segmentation CNN made of a series of convolutions and deconvolutions. The convolution feature maps are connected to the deconv maps of the same size. The network was tested on the 2 class 2D ISBI cell segmentation dataset. Used the crossentropy loss and a lot of data augmentation.