.png){kind=icon}
.png){kind=icon}
Return
Detecting Learning Behaviors Using Deep-learning-based Classroom Teaching Quality Analysis System
.png){kind=icon}
PRE
AI
DOI：10.18494/SAM6063.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The application of the deep-learning-based You Only Look Once (YOLO) network model for identifying teaching and learning behaviors in classrooms has attracted attention in recent studies. However, there is limited scholarly focus on using these models to evaluate instructional quality. In this study, we propose a system to analyze instructional quality through advanced deep learning technology, aiming to assess and enhance university classroom activities. It comprises two main components. First, the YOLOv8 architecture supports the classroom teaching quality analysis network (CTQA-YOLOv8-NET), which detects and categorizes student behaviors into six types during class. The second component is a visual platform for analyzing teaching quality based on data obtained through CTQA-YOLOv8-NET. By extracting one image frame per minute from a 45-minute course video, this method visually represents the distribution of student behaviors across categories and compiles data into trend graphs showing effective versus ineffective learning ratios over time. These visual tools provide scientific evidence for evaluating instructional effectiveness, allowing teachers to objectively assess outcomes and develop targeted improvement strategies in areas such as content adjustments and pedagogical innovations. In this study, these methods were specifically applied to the university general educational class, Ideology, Morality, and Rule of Law, suggesting optimization options while examining the strengths and weaknesses of the course, ultimately confirming the effectiveness of interventions in enhancing instructional quality.
Keywords:
YOLOv8 network
classroom behavior detection
classroom teaching quality analysis system
CTQA-YOLOv8-NET