Detecting students’ classroom behaviors from instructional videos is important for instructional assessment, analyzing students’ learning status, and improving teaching quality. To achieve effective detection of student classroom behavior based on videos, this paper proposes a classroom behavior detection model based on the improved SlowFast. First, a Multi-scale Spatial-Temporal Attention (MSTA) module is added to SlowFast to improve the ability of the model to extract multi-scale spatial and temporal information in the feature maps. Second, Efficient Temporal Attention (ETA) is introduced to make the model more focused on the salient features of the behavior in the temporal domain. Finally, a spatio-temporal-oriented student classroom behavior dataset is constructed. The experimental results show that, compared with SlowFast, our proposed MSTA-SlowFast has a better detection performance with mean average precision (mAP) improvement of 5.63% on the self-made classroom behavior detection dataset.

Face detection in the classroom environment is the basis for student face recognition, sensorless attendance, and concentration analysis. Due to equipment, lighting, and the uncontrollability of students in an unconstrained environment, images include many moving faces, occluded faces, and extremely small faces in a classroom environment. Since the image sent to the detector will be resized to a smaller size, the face information extracted by the detector is very limited. This seriously affects the accuracy of face detection. Therefore, this paper proposes an adaptive fusion-based YOLOv5 method for face detection in classroom environments. First, a very small face detection layer in YOLOv5 is added to enhance the YOLOv5 baseline, and an adaptive fusion backbone network based on multi-scale features is proposed, which has the ability to feature fusion and rich feature information. Second, the adaptive spatial feature fusion strategy is applied to the network, considering the face location information and semantic information. Finally, a face dataset Classroom-Face in the classroom environment is creatively proposed, and it is verified with our method. The experimental results show that, compared with YOLOv5 or other traditional algorithms, our algorithm portrays better performance in WIDER-FACE Dataset and Classroom-Face dataset.