Multi-student individual segmentation and state autonomous identification method based on deep learning

Abstract

The invention provides a multi-student individual segmentation and state autonomous recognition method based on deep learning, and the method comprises the steps: collecting a normal class video, carrying out the framing, and obtaining continuous frame images of a class video; segmenting out students' individuals and non-students' individuals, and marking the non-students' individuals as masks ofdifferent colors; finding a human eye feature point of each student individual through a human face feature point positioning method, calculating a human eye closure feature value of each student individual, and judging whether each student individual is in a lecture attending state at present or not; judging whether faces are detected by all the student individuals or not, and judging the class attending state grades of the student individuals; and designing the student lecture attending efficiency evaluation method in the whole class period according to whether each student is in a lecture attending state or not and whether each student is in a non-head-up state or not. The invention provides a solution for realizing the autonomous identification of the student lecture attending state and the judgment of the lecture attending efficiency, and has the advantages of high speed, high identification rate and strong environmental adaptability.

Description

technical field [0001] The invention relates to the technical field of machine learning and machine vision, in particular to a method for individual segmentation of multiple students and autonomous state recognition based on deep learning. Background technique [0002] In recent years, the concept of smart classroom has emerged. It is an emerging concept that efficiently integrates advanced information collection and transmission technology, various intelligent sensing technologies, and computer processing technology into the education field. The classroom is the core of education, and the status of students attending classes is the most direct and effective reflection of classroom efficiency. The existing classroom teaching feedback is still in the subjective manual analysis stage, which consumes energy, is inefficient, and cannot comprehensively monitor each student's performance. state. Target segmentation is the basis of visual analysis. Convolutional neural networks ar...

AI Technical Summary

Problems solved by technology

[0003] At present, there are few methods for identifying students' listening status. There are methods based on human body feature point recognition and pressure sensor data collection. These methods still have some inevitable defects, such as strong subjectivity, low accuracy, and high cost.

Images