{%hackmd SybccZ6XD %} ###### tags: `paper` # U-Net: Convolutional Networks for Biomedical Image Segmentation Goal > Segmentation Motivaction > fully convolutional network [9] Architecture > ![](https://hackmd.io/_uploads/S17Y7CNE3.png) How to combine the blue and white part in the right side? > Concat Why they crop the image and input to the right side? > Due to unpadding convolution What is up-convolution? > https://medium.com/ai%E5%8F%8D%E6%96%97%E5%9F%8E/%E5%8F%8D%E6%8D%B2%E7%A9%8D-deconvolution-%E4%B8%8A%E6%8E%A1%E6%A8%A3-unsampling-%E8%88%87%E4%B8%8A%E6%B1%A0%E5%8C%96-unpooling-%E5%B7%AE%E7%95%B0-feee4db49a00