Learning State Classification And Practice Based On Video

With the rapid development of the internet and the younger audience for intelligent products such as mobile phones and i Pads,it is easy for younger students to be distracted from their studies.How to improve the concentration of younger students when they are studying alone has become an issue that has been studied by education experts.In this paper,we develop a tomato clock system based on learning states that can help improve concentration.We propose a joint model to detect students’ learning states and thus control whether the tasks of the tomato clock can be completed properly,which in turn motivates students to focus on their studies.The main work of this paper includes.（1）The first is to shoot videos in real time while students are studying.The spatiotemporal action detection model Slow Fast is used to detect the actions of the students in the video,and the learning state is preliminarily classified by actions.（2）When the students do not show behaviors and actions and are in an approximate stationary state,the Slow Fast model can only detect the action category of "sitting".By detecting the key points of the students’ faces,we calculate the students’ head posture information and determine their gaze.The area is the field of view.Then use the FPN-Faster RCNN model to detect representative key items such as mobile phones,books,etc.that appear within the students’ field of vision,and then identify learning states such as "reading" and "looking at mobile phones".（3）Apply the above method of detecting students’ learning status to the mobile terminal,use the flutter technical framework to develop the mobile terminal Pomodoro App,call the camera of the mobile terminal device to record the video,upload it to the server for detection,and judge the learning status of the students in the video,so as to help Judging whether the Pomodoro task has been completed allows parents to supervise students’ learning and help students improve their learning concentration.In this paper,it is proved by experiments that the method in this paper can achieve an accuracy of 85.28% for the classification of the learning state,and can be well applied to the Pomodoro system.