An Automatic Scoring Algorithm for Physics Experiments Based on YOLOv5 and MobileNet

The accurate evaluation of students' physics experiment operation process is of great significance for the middle school entrance examination. In order to achieve automatic scoring of experimental operations. This study proposes a cascaded network model combining YOLOv5 and MobileNet for automatic detection of experimental equipment status. The model first uses an improved YOLOv5 for device localization and preliminary recognition. The improved YOLOv5 network model has an average accuracy improvement of 3.6% and a parameter reduction of 11.0%. Subsequently, MobileNet was selected to perform fine-grained classification of the experimental equipment status. Finally, by integrating the outputs of YOLOv5 and MobileNet, the model achieved accurate detection of experimental equipment status, providing effective technical support for automatic scoring systems.