The deep learning method requires a lot of data to achieve good performance, but suffers from insufficient of data. This problem is particularly remarkable in the medical image processing. To solve this problem, a method of weakly supervised learning using pseudo-label is widely used. In this paper, we propose a method of segmenting the vertebral body by augmentation the data using the weakly supervised method in the lateral X-ray image of infants. The proposed method is designed based on U-Net network, which is widely used in medical image segmentation problems, and consists of one encoder and two decoders. Experimentally, the result shows that the performance of our method has been improved by 1.05 % over the previous method.