A live course evaluation method based on deep learning

Abstract

The present invention relates to a live course evaluation method based on deep learning, which specifically includes the following steps: S1, obtaining the student's facial expression history data, training the recognition of the facial expression history data according to deep learning, and obtaining a classroom expression recognition model; S2, the camera at the student end collects the student's facial expression data, and the classroom expression recognition model recognizes and obtains learning status information; S3, sending the learning status information to the server, and the server outputs the overall understanding degree of the current knowledge point after real-time aggregation and analysis as the live course evaluation result. Compared with the prior art, the present invention has the advantages of saving a large amount of network resources, reducing the transmission pressure of the server, reducing the model accuracy as little as possible while reducing the model size, and having high timeliness.

Description

Technical Field

[0001] The present invention relates to the field of online education technology, and in particular to a live course evaluation method based on deep learning. Background Art

[0002] We are now entering an era of universal learning. Online classes are not just a slogan, but have gradually become a new form of teaching. When people can only study at home, in order to carry out teaching tasks normally, most colleges and universities have adopted online teaching methods. Among them, online live classroom learning has its own advantages compared to traditional face-to-face classrooms, and is the teaching method chosen by most college teachers. In the current teaching methods, there are several difficulties in collecting students' evaluations or feedback:

[0003] 1. If the number of students is relatively small, all students can be on the microphone, and the interaction between teachers and students is relatively real-time, but it will generate a lot of traffic, consume a lot of resources, and the income generated is not very high. Most of the time, teachers just observe students' expressions and judge their learning situation. In order to see students' expressions, a large number of image resources must be uploaded, which will inevitably cause a certain amount of resource waste and network congestion. It is bound to face the problem of high cost and difficulty in promotion;

[0004] Second, if there are a lot of students in class, we cannot let everyone take the microphone. If the teacher wants to know the situation of each student, most of the time he can only manually input the situation through the chat box. The teacher has to pay attention to the chat box at all times, which is easy to be distracted, and there are a lot of spam messages that will take up the teacher's time. At the same time, not paying attention in time will also affect the quality of students' learning;

[0005] 3. Alternatively, students may choose to submit feedback at the last minute. However, due to the lack of timely feedback or oversimplification of the evaluation, the collected evaluation cannot fully reflect the teaching quality. In addition, the delay in after-class feedback will also cause a gap in learning. Students may not remember the questions they had when reviewing the lesson again, which increases the cost of learning. Summary of the invention

[0006] The purpose of the present invention is to provide a live course evaluation method based on deep learning in order to overcome the defects of the above-mentioned prior art that live teaching lacks a suitable evaluation system or that a large amount of cost is required to obtain a good evaluation.

[0007] The purpose of the present invention can be achieved by the following technical solutions:

[0008] A live course evaluation method based on deep learning specifically includes the following steps:

[0009] S1. Obtain the students' facial expression history data, train the recognition of the facial expression history data based on deep learning, and obtain a classroom expression recognition model;

[0010] S2, the camera at the student end collects the student's facial expression data, which is recognized by the classroom expression recognition model to obtain learning status information;

[0011] S3. Send the learning status information to the server. The server summarizes and analyzes the information in real time and outputs the overall understanding level of the current knowledge points as the evaluation result of the live course.

[0012] The facial expression history data in step S1 is specifically the facial expression data of the student obtained in an offline state.

[0013] Furthermore, the facial expression data of the trainees acquired in the offline state is trained by the server in the offline state.

[0014] Furthermore, the server trains the classroom expression recognition model by means of model distillation.

[0015] The recognition content of the facial expression history data during training in step S1 includes whether the student is listening to a lecture and the student's learning status.

[0016] The step S2 also includes the student obtaining a classroom expression recognition model from the server.

[0017] In step S2, the student terminal recognizes facial expression data through local client computing resources.

[0018] The type of the learning status information in step S2 is text information.

[0019] The step S3 also includes the server displaying the learning status information of all current students on the teacher's side.

[0020] Furthermore, the teacher terminal and the corresponding multiple student terminals are respectively connected to the server and perform data communication through the server.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. The present invention adopts the client deployment mode and only uploads the recognition results, which will save a lot of network resources and reduce the data transmission pressure of the server. At the same time, there is no picture upload, which can better protect the privacy of students locally; and the consumed computing power is apportioned to each client, thereby reducing the pressure on the server and the pressure on each client is relatively small.

[0023] 2. Since different models have different efficiencies and the model needs to be deployed on the client, it is bound to have certain requirements on the size and accuracy of the model. The present invention makes full use of the model distillation method to reduce the model accuracy as little as possible while reducing the model size, thereby ensuring the accuracy and stability of model recognition.

[0024] 3. The present invention improves the live broadcast evaluation system, makes up for the shortcomings of the previous single text evaluation method, and has high timeliness; at the same time, the trained model has a certain versatility and can be used in similar scenarios and in other learning live broadcasts. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0026] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0027] Example

[0028] like Figure 1 As shown, a live course evaluation method based on deep learning specifically includes the following steps:

[0029] S1. Obtain the students' facial expression history data, train the recognition of the facial expression history data based on deep learning, and obtain a classroom expression recognition model;

[0030] S2, the camera at the student end collects the student's facial expression data, which is recognized by the classroom expression recognition model to obtain learning status information;

[0031] S3. Send the learning status information to the server. The server summarizes and analyzes the information in real time and outputs the overall understanding level of the current knowledge points as the evaluation result of the live course.

[0032] The facial expression history data in step S1 is specifically the facial expression data of the student obtained in an offline state.

[0033] The facial expression data of the students obtained in an offline state is trained by the server in an offline state.

[0034] The server trains the classroom expression recognition model through model distillation.

[0035] In step S1, the recognition content of the facial expression history data during training includes whether the student is listening to a lecture and the student's learning status.

[0036] Step S2 also includes the student obtaining a classroom expression recognition model from the server.

[0037] In step S2, the student terminal recognizes facial expression data through local client computing resources.

[0038] The type of the learning status information in step S2 is text information.

[0039] Step S3 also includes the server displaying the learning status information of all current students on the teacher's side. In this embodiment, the learning status information of all students is displayed on a real-time large screen.

[0040] The teacher's terminal and corresponding multiple student terminals are connected to the server respectively, and data communication is carried out through the server.

[0041] In the specific implementation, based on the learning status information of all current students displayed on the teacher's side, the teacher can provide additional explanations on the points that students have more questions about. At the same time, in response to the teacher's questions, students can also give the most timely feedback through simple expressions.

[0042] In addition, it should be noted that the specific embodiments described in this specification may be named differently, and the above contents described in this specification are merely examples of the structure of the present invention. Any equivalent changes or simple changes made based on the structure, features and principles of the present invention are included in the protection scope of the present invention. Those skilled in the art of the present invention may make various modifications or supplements to the specific examples described or adopt similar methods, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A live course evaluation method based on deep learning, characterized in that: The specific steps include: S1. Obtain the student's facial expression history data, train the recognition of the facial expression history data according to deep learning, and obtain a classroom expression recognition model. The facial expression history data is specifically the student's facial expression data obtained in an offline state. The student's facial expression data obtained in an offline state is trained by the server in an offline state. The recognition content of the facial expression history data during training includes whether the student is listening to the lecture and the student's learning status. The server trains the classroom expression recognition model by model distillation. S2, the camera at the student end collects the student's facial expression data, and the classroom expression recognition model recognizes the data to obtain the learning status information. The type of the learning status information is text information. The student end obtains the classroom expression recognition model from the server, and the student end realizes the recognition of facial expression data through the local client computing resources; S3. Send the learning status information to the server. The server summarizes and analyzes the information in real time and outputs the overall understanding level of the current knowledge points as the evaluation result of the live course.

2. According to a method for evaluating live courses based on deep learning according to claim 1, it is characterized in that: The step S3 also includes the server displaying the learning status information of all current students on the teacher's side.

3. A live course evaluation method based on deep learning according to claim 2, characterized in that: The teacher terminal and the corresponding multiple student terminals are respectively connected to the server and perform data communication through the server.

Images