在智慧教育体系快速发展进程中，传统课堂观测手段存在的效能缺陷与主观偏差问题逐渐显现。针对上述问题，本研究提出一种基于学生行为检测的学生课堂专注度评估模型，其技术实现框架大致包含三个模块。在模型选择阶段，通过系统对比单双阶段目标检测模型的性能差异，重点分析YOLOv8与YOLOv11的架构特性，并确定以YOLOv8作为基础框架。在此基础框架下，引入全局注意力机制（GAM）优化特征提取过程，形成增强型目标检测模型YOLOv8-GAM。基于行为检测，研究构建动态权重分配算法与专注度量化评估模型，综合行为频次与检测置信度参数进行分值计算，配套开发的可视化分析可实时呈现专注度动态变化曲线、预警信号以及专注度评估报告。该课堂专注度评估系统为教育管理者提供了过程性评估工具，助力教师实施精准教学干预，推动教育评估方法向过程化与智能化转型。

With the rapid development of intelligent education systems, traditional classroom observation methods increasingly reveal limitations in efficiency and objectivity. To address these issues, this study proposes a student classroom engagement assessment model based on behavior detection. The technical framework comprises three key components: (1) Model selection: Through systematic comparison of single-stage and dual-stage object detection models, YOLOv8 was selected as the baseline architecture after analyzing its structural advantages over YOLOv11. (2) Model optimization: The Global Attention Mechanism (GAM) was integrated into YOLOv8 to enhance feature extraction capabilities, resulting in an improved YOLOv8-GAM model. (3) System implementation: A dynamic weight allocation algorithm and quantitative engagement assessment model were developed, combining behavioral frequency and detection confidence parameters to calculate engagement scores. A visualization platform was designed to display real-time engagement dynamics, generate alerts, and produce evaluation reports. This system provides educators with process-oriented assessment tools to support instructional interventions, promoting the transformation of educational evaluation towards process-oriented and intelligent paradigms.