Currently, the dataset of children’s teeth suffers from the problems of less data, expensive annotation, poor segmentation of the edges of the teeth, and difficulty in accurately segmenting the detailed information of the edges of the teeth. In this paper, we design a deep learning algorithm named UX-CNet that can effectively segment teeth and solve the problem of poor effect of teeth edge segmentation. In the experiment, data augmentation was performed on the children’s teeth image dataset to improve the learning ability of the model, followed by training the data with UX-CNet to improve the model in terms of overall segmentation and detail segmentation. Our method achieved an average DSC score of 95.64% and an average HD error of 0.03 for the teeth segmentation on the validation. The experiments show that UX-CNet has a great improvement over the baseline algorithm and other algorithms, proving the effectiveness of the method.