随着互联网技术的发展，在线教学的可行性和实用性得到验证。作为教育领域的重要变革，在线教学能够弥补常规教学在学习时间、地点和可复现性等方面的不足。尤其在2019年爆发的全球性新冠疫情中，我国教育部提出高校及中小学应利用教学资源开展在线教学，实现“停课不停学”。及时保障了广大学生和教师群体的学习工作秩序。但是在线教学由于空间限制，教师无法像传统教学那样掌握学生学习状态。

针对该问题，本文基于OpenFace2.0面部行为分析工具，对学生在线学习视频中的头部姿态和面部标志点进行检测。并根据所得眼动、凝视方向估计、口部、头部姿态估计等数据，计算出学生的专注度和疲劳度，进而计算出学生的在线学习投入度，实现了一种在线学习状态评价体系。经过与人工评价对比，本方法具有很好的准确性。能够在实际的在线教学中，为教师的学情分析、教学监督、教学反思以及对学生学习的过程性评价提供新思路，具有较好的实际应用价值。

With the development of Internet technology, the feasibility and practicability of online teaching have been verified. As an important change in the field of education, online teaching can make up for the shortcomings of conventional teaching in terms of learning time, place and reproducibility. Especially in the global COVID-19, China's Ministry of Education proposed that universities, primary and secondary schools should use teaching resources to carry out online teaching, to achieve "classes without suspension". In a timely manner to ensure the majority of students and teachers learning and working order. However, due to the space limitation of online teaching, teachers cannot master students' learning status like traditional teaching.

To solve this problem, based on the OpenFace2.0 facial behavior analysis tool, this paper completes the detection of head poses and facial landmarks in students' online learning videos. According to the data of eye movement, gaze direction estimation, mouth and head posture estimation, the students' concentration and fatigue degree were calculated, and then the students' online learning engagement degree was calculated, realizing an online learning state evaluation system. Compared with manual evaluation, this method has good accuracy. In the actual online teaching, it can provide new ideas for teachers' learning situation analysis, teaching supervision, teaching reflection and process evaluation of students' learning, and has good practical application value.