Many deep learning models are capable of performing various image segmentation tasks with excellent performance on well-annotated datasets. 2D panoramic dental image segmentation, as a specific task within the field of medical image segmentation, plays a crucial role in providing reference material for auxiliary diagnosis or subsequent downstream tasks. However, the reality is that there is a scarcity of well-annotated related datasets in this field, making the development of accurate and efficient semi-supervised dental image segmentation methods a considerable challenge. This paper presents a teeth segmentation method based on a semi-supervised segmentation research via cross teaching between CNN and Transformer. And we designed and used some pre-processing and post-processing methods to complete the migration of the original research on the 2D dental panoramic X-ray tooth images segmentation task. Finally, Our method achieved an average instance Dice score of 87.07% and average instance NSD score of 41.08% for the teeth segmentation on the validation set and instance using a NVIDIA GeForce RTX 4090. The average running time was 2.36 s for one image. The code is available at https://github.com/aicorein/STS2024-Semi-Supervised-Cross-Teaching .