CBPH-Net: A Small Object Detector for Behavior Recognition in Classroom Scenarios

Recognizing classroom behavior is crucial for assessing and improving teaching quality. However, the existing methods for behavior recognition have limited accuracy due to issues, such as occlusions, pose variations, and inconsistent target scales. To address these challenges, we propose an advanced single-stage object detector called ConvNeXt Block Prediction Head Network (CBPH-Net). Specifically, we design an efficient feature extraction module (FEM) to capture more channel information and relevant features from the images in the backbone network. The neck network combines the path aggregation network (PANet) architecture and coordinate attention (CA) to integrate semantic and positional information and suppress irrelevant background information, enabling the network to accurately locate students. CBPH utilizes convolutional kernels of different sizes and parsing multiscale features to enhance the multiscale recognition capability of CBPH-Net especially for accurate detection of small objects. To reduce the influence of irrelevant background, we use elliptical boxes instead of rectangular boxes when calculating the similarity between ground-truth and predicted values. In addition, we construct a dataset named Student–Teacher Behavior Dataset (STBD-08) that contains 4432 images with 151574 labeled anchors covering eight typical classroom behaviors. On the proposed dataset STBD-08, CBPH-Net achieves a mean average precision (mAP) of 87.5% (an improvement of 3.4% compared with YOLOv5 and 1.2% compared with YOLOv7). It processes one frame with the latency of 31.3 ms (1 ms slower than YOLOv5 and 5.3 ms faster than YOLOv7). Moreover, it achieves a precision of 75.7% in small object recognition, surpassing all comparative methods. The experimental results demonstrate that the CBPH-Net can be efficiently applied to classroom behavior recognition tasks. Codes and datasets are available at
https://github.com/icedle/CBPH-Net
.