A training method for face model of class students

Abstract

The invention discloses a training method for face models of students in a class. Firstly, import a specified class seating chart and the student status photo of each student in the class; use the face detection technology Mtcnn to identify the identity of each student. Face activity area frame; establish the mapping relationship between the human face activity area frame and the position sequence in the seating table to form a human face activity area frame with coordinate mapping; use artificial aids to detect people with wrong positions and those who do not meet the requirements Manually adjust the face activity area frame; use face detection technology to generate batches of student face photo samples from screenshots of student panoramic videos, and file them under each student's name and student status photo; eliminate false detection and false filing The student face photo samples; and then use the verified student face photo samples to train the face model. The above method can liberate people from tedious labor and improve the efficiency of face model training for class students.

Description

technical field [0001] The invention relates to the technical field of face model training, in particular to a training method for class student face models. Background technique [0002] In the field of artificial intelligence, when the network model is established, it often requires a large number of data samples for training to achieve the desired effect. Similarly, when artificial intelligence technology is applied to classroom teaching, it often requires a large number of class data samples for training in order to more accurately identify each student's posture behavior and facial features. In the early stage of face model training, it is usually necessary to manually calibrate the identified student sample data to increase the recognition accuracy. [0003] The student model training in the class has different characteristics from the face model training in the general scene. First, the distance between the students and the camera is different. There will be situati...

AI Technical Summary

Problems solved by technology

[0003] The student model training in the class has different characteristics from the face model training in the general scene. First, the distance between the students and the camera is different. There will be situations where only half of the head of the students in the back row is missing; the second is that the students in the same class belong to the same age group in terms of physical development, and their hairstyles, skin colors, clothing, and expressions have strong similarities, and it is not easy to provide background information support

Accurate face recognition requires a large number of face sample photos. If the method of manual calibration of student samples is used for model training, not only the work is cumbersome and boring, but also the training will consume more time and energy.

Images