Dentition identification and root canal are very common in dentistry. Dentition identification helps dentists to diagnose the teeth conditions of patient and discuss about their treatments, while root canal is one of the common treatments. The goal of this paper is to label tooth dentition and identify root canal via GAN and rule-based algorithm. Before inputting images into GAN, the original images were sharpened to make the learning task easier. Next, we use the rule-based algorithm to identify root canal. For evaluating proposed method, we invite three dentists to evaluate the accuracy of the results, and they consider that the results are reliable. We also discuss the impact that iterations have on accuracy. The model becomes stable when iteration comes to 12000, and the accuracy reaches 93.7%.