[深度学习论文笔记]3D U-Net: Learning Dense Volumetric Segmentation from Sparse Annotation 从稀疏标注学习密集体分割