In China, training schools are ubiquitous, where heterogeneous pedestrian flow (which consists of adults and children) widely occurs during the after-class period. In this paper, we develop a fine grid cellular CA model to describe the pedestrian behaviors (e.g., pick-up behavior, searching behavior, matching behavior, waiting behavior, leading behavior, and following behavior) at a training school during the after-class period and explore the effects of the special behaviors on each pedestrian’s movement in and around a classroom. To describe the heterogeneous pedestrian flow accurately, (i) some questionnaire surveys are designed to extract some features of adult’s and child’s movement, and (ii) some video experiments are conducted to estimate/calibrate some parameters of interest in the proposed model. Finally, some strategies are designed to enhance the evacuation efficiency and the operational efficiency of training school. The numerical results indicate that the proposed model can reasonably match with reality, and the proposed strategies can enhance the evacuation efficiency and the operational efficiency of training school. The results can help the administrators to effectively manage the pedestrian evacuation at training school during the after-class period.