Examination of Abnormal Behavior Detection Based on Improved YOLOv3

Examination is a way to select talents, and perfect invigilation strategy can improve the fairness of examination. To realize automatic detection abnormal behavior in examination room, method based on improved YOLOv3 (The third version You Only Look Once algorithm) algorithm proposed. The by using K-Means algorithm, GIoUloss, focal loss, Darknet32. In addition, frame-alternate dual-thread used optimize process. research results show that both accuracy speed. greatly increase mean Average Precision (mAP) test set reached 88.53%, speed 42 Frames Per Second (FPS) method. provide certain reference for automated invigilation.