The primary school stage is the first step for children to enter the school life, which plays a crucial role in the subsequent academic development. This paper studies the teacher role influence based on Generative Adversarial Network. We use generative adversarial networks to deal with complex data. A fusion data enhancement scheme is proposed to make the generator fit the distribution of actual samples by training the generation adversarial network and generating samples approximating the distribution by the input random noise. Combined with the traditional data enhancement method of image flipping, the enhancement algorithm can enrich the diversity of samples, balance the distribution of samples, and expand the distribution range of samples in the feature space, which is conducive to improving the algorithm's performance.