Student Break Behavior Recognition Dataset

With the explosive growth of classroom video data, it is possible to utilize artificial intelligence technology to recognize students' break behavior, which is very beneficial for personalized teaching. However, the lack of a dataset specifically for student break behavior may prevent its development. Therefore, this paper constructs a dataset of student break behavior. The dataset collected 64 break videos from Beijing Normal University. After processing, 1,538 video samples containing 9 student break behavior categories were obtained. Because we collect real video data, the student behavior is spontaneous, which makes the dataset representative and realistic. In addition, we also analyzed the dataset. Finally, it provides a baseline for the dataset and compares it with the current mainstream dataset. The results show that our dataset is feasible and reliable.