Abstract—A new teacher at an elementary school, junior high school, or a high school has to teach classes from the first day. However, new teachers often find teaching difficult in a real-world teaching environment. In this paper, we analyze the gestures and behaviors of a beginner teacher and an expert teacher with the aim of increasing the teaching quality of beginner teachers. Image processing technology was used to automatically report visualized results of teaching behaviors. We devised the following approach: (1) a new teacher conducts a class in an environment replicating a real class, (2) the class behavior of the new teacher is systematically evaluated, and (3) new teachers objectively look back at their classes and derive insights for their development. The aim of this research is to evaluate and visualize the behavior of teachers during classes, focusing on (2) and (3). Specifically, we took videos of a trial class of expert and beginner teachers using a video camera (third-person view) and egocentric vision. The egocentric vision recognizes objects using YOLO algorithm, while the third-person view classifies the teacher behavior using Spatial-Temporal Graph Convolution Networks (ST-GCN) based on Open Pose. Then, we analyzed the differences between expert teacher behavior and beginner teacher behavior and visualized the results. In the behavioral analysis at STGCN, ‘Writing on the board’ constituted 74.5% of the approach of beginner teachers and ‘Pointing to the board’ was 36.9%, whereas ‘Writing on the board’ was 33.1% of the approach of expert teachers. Further, in the behavioral analysis by the YOLO algorithm, ‘Writing on the board’ was 41.1% and ‘Talking at the front’ 57.8% for beginner teachers, and ‘Writing on the board’ was 25.3% and ‘Talking at the front’ 74.0% for expert teachers. In other words, we confirmed that the experts were conscious of the whole classroom, and that beginners tended to do lessons only by writing on the board.