Classroom attendance anomaly detection method based on illumination generation antagonism network

Abstract

The invention discloses a class attendance abnormal detection method based on illumination generation antagonism network, includes collecting real class head posture data, rendering illumination classhead posture data, building illumination generation antagonism network, generating antagonism samples, building head posture detection model, class head posture detection, class attendance abnormal detection. By using the depth neural network, the invention improves the positioning accuracy of the head area and reduces the interference of the non-head area to the judgment of the non-attending state.

Description

technical field [0001] The invention belongs to the technical field of anomaly detection, and more specifically, relates to an anomaly detection method for classroom listening based on an illumination generation confrontation network. Background technique [0002] Computer anomaly detection is to use computer vision theory and video analysis methods to analyze the video sequences recorded by cameras and other monitoring equipment without human intervention, so as to realize the positioning, identification and tracking of targets in structured scenes, and here Based on the analysis and judgment of the target's behavior, the understanding of the meaning of the image content and the interpretation of the objective scene are obtained, and thus guide and plan actions. [0003] Existing anomaly detection methods often use specific statistical analysis methods and deep learning methods. The Chinese patent with application number 201510141935.6 "A method for abnormal detection of c...

AI Technical Summary

Problems solved by technology

However, the method is complex, the scene adaptability is low, a large number of real data sets are required, and the data cost is large.

[0005] Existing anomaly detection methods have not yet involved the analysis of listening status in classroom teaching invigilation process, the task of the present invention is different from existing classroom video, head posture data and existing anomaly detection methods

Images