The study uses convolutional neural network in the teaching practice of “Microcontroller Principles and Applications” course, detects and recognizes students’ expression and recognition in the classroom through lightweight convolutional neural network and two-stream neural network, respectively, and gives real-time feedback on the detection and recognition results to construct an instant feedback model of classroom teaching effect. The teaching experiment is designed around the teaching model based on convolutional neural network, and the practical effect of convolutional neural network in the course of Microcontroller Principles and Applications is explored from the classroom emotional activeness, students’ head-up rate, and course grades of Microcontroller Principles and Applications before and after the teaching experiment. The emotional activeness of the “Microcontroller Principles and Applications” classroom at the beginning of the semester in the middle of the semester and at the end of the semester was average, good, and excellent, respectively, and the students’ head-up rate was 37.26%, 59.85%, and 85.31%, respectively. Emotional activity and students’ head-up rate improved as the teaching experiment progressed. There was no significant difference between the performance of the experimental group and the control group in Principles and Applications of Microcontrollers before the experiment, and the P-values were all greater than 0.5. After the teaching experiment, the P-values of the five dimensions were all less than 0.5, and a large gap between the performance of the two groups emerged. The experimental group’s posttest scores were significantly higher than those of the control group.