The document presents a new deep learning model called BATA-Unet for liver segmentation. BATA-Unet is based on the Unet architecture but adds batch normalization layers after each convolution layer. The model was tested on two datasets, MICCAI and 3D-IRCAD, achieving Dice scores of 0.97 and 0.96 respectively. This outperforms the authors' previous BATA-Convnet model as well as other state-of-the-art models for liver segmentation. The document provides background on liver segmentation and reviews several related works that use deep learning and other techniques for medical image segmentation.