CN-115984642-B - Data enhancement method, device, equipment and storage medium

Abstract

The application is applicable to the technical field of data processing and provides a data enhancement method, a device, equipment and a storage medium, wherein the data enhancement method comprises the steps of respectively extracting features of image data in a target data set and a training data set to obtain a target data feature vector and a training data feature vector; the method comprises the steps of carrying out distinguishing feature fusion on the target data feature vector and the training data feature vector to obtain a fusion feature vector, generating fusion image data corresponding to the fusion feature vector according to the fusion feature vector, and updating the image data in the training data set into the fusion image data. The application can fuse the distinguishing characteristics of the individual difference data appearing in the target data set to the image data in the training data set, so that the image data in the enhanced training data set can also cover the individual difference data.

Inventors

• YAN RUIHAI

Assignees

• 大连熵基科技有限公司

Dates

Publication Date

20260505

Application Date

20221209

Claims (9)

1. 1. A method of data enhancement, comprising: Respectively extracting features of image data in a target data set and a training data set to obtain a target data feature vector and a training data feature vector; Performing distinguishing feature fusion on the target data feature vector and the training data feature vector to obtain a fusion feature vector; Generating fusion image data corresponding to the fusion feature vector according to the fusion feature vector; updating the image data in the training data set to the fused image data; The distinguishing feature fusion is performed on the target data feature vector and the training data feature vector to obtain a fusion feature vector, which is expressed as follows: F' x =F x +z(F t -F x ) wherein z represents a linear orthogonal basis subspace for generating a countermeasure network, and is used for controlling the fusion degree of distinguishing features F t -F x , F t is the target data feature vector, F x is the training data feature vector, and F' x is the fusion feature vector.
2. 2. The data enhancement method according to claim 1, wherein the feature extraction is performed on the image data in the target data set and the training data set, respectively, to obtain a target data feature vector and a training data feature vector, and the method comprises: And respectively carrying out feature extraction on the image data in the target data set and the training data set through a feature extraction network to obtain a target data feature vector and a training data feature vector.
3. 3. The data enhancement method according to claim 1, wherein the generating, from the fusion feature vector, the fusion image data corresponding to the fusion feature vector includes: And generating fused image data corresponding to the fused feature vector through a generating network according to the fused feature vector.
4. 4. A data enhancement method according to any one of claims 1 to 3, wherein said updating the image data in the training dataset to the fused image data comprises: calculating the loss corresponding to the fused image data according to the fused image data; and when the loss corresponding to the fused image data is smaller than a preset threshold value, updating the image data in the training data set into the fused image data.
5. 5. The data enhancement method of claim 4, wherein said calculating a corresponding loss of said fused image data from said fused image data comprises: According to the fusion image data, respectively calculating corresponding cross entropy loss, antagonism loss and discrimination loss of the fusion image data; and calculating the sum of the cross entropy loss, the antagonism loss and the orthogonality loss of the orthogonal base subspace to obtain a generated network loss.
6. 6. The data enhancement method according to claim 4, wherein updating the image data in the training dataset to the fused image data when the loss corresponding to the fused image data is less than a preset threshold value comprises: And when the generated network loss corresponding to the fused image data is smaller than a first preset threshold value and the distinguishing loss corresponding to the fused image data is smaller than a second preset threshold value, updating the image data in the training data set into the fused image data.
7. 7. A data enhancement device, comprising: The feature extraction module is used for extracting features of the image data in the target data set and the training data set respectively to obtain a target data feature vector and a training data feature vector; the feature fusion module is used for carrying out distinguishing feature fusion on the target data feature vector and the training data feature vector to obtain a fusion feature vector; The image generation module is used for generating fusion image data corresponding to the fusion feature vector according to the fusion feature vector; a data updating module, configured to update the image data in the training data set to the fused image data; The distinguishing feature fusion is performed on the target data feature vector and the training data feature vector to obtain a fusion feature vector, which is expressed as follows: F' x =F x +z(F t -F x ) wherein z represents a linear orthogonal basis subspace for generating a countermeasure network, and is used for controlling the fusion degree of distinguishing features F t -F x , F t is the target data feature vector, F x is the training data feature vector, and F' x is the fusion feature vector.
8. 8. A terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the data enhancement method according to any of claims 1 to 6 when executing the computer program.
9. 9. A computer readable storage medium storing a computer program, characterized in that the computer program, when executed by a processor, implements the data enhancement method according to any one of claims 1 to 6.

Description

Data enhancement method, device, equipment and storage medium Technical Field The present application relates to the field of data processing technologies, and in particular, to a data enhancement method, device, apparatus, and storage medium. Background In the face image recognition by the deep learning model, since the acquired face image data has a variety, it is impossible for the training data set of the deep learning model to encompass all types of face image data. The deep learning model trained by the training data set can encounter a small amount of difference data with distinguishing characteristics when applied, and when the training data set lacks the data, the deep learning model is easy to identify errors. To solve this problem, data enhancement is required for the training data set, and a general method is to perform data enhancement in a random direction on the training data set, but it is difficult to cover individual difference data that may occur. Disclosure of Invention The embodiment of the application provides a data enhancement method, a device, equipment and a storage medium, which can solve the problem that the training data enhancement direction is random and the possible individual difference data is difficult to cover in the prior art. A first aspect of an embodiment of the present application provides a data enhancement method, including: Respectively extracting features of image data in a target data set and a training data set to obtain a target data feature vector and a training data feature vector; Performing distinguishing feature fusion on the target data feature vector and the training data feature vector to obtain a fusion feature vector; Generating fusion image data corresponding to the fusion feature vector according to the fusion feature vector; And updating the image data in the training data set into the fused image data. A second aspect of an embodiment of the present application provides a data enhancement device, including: The feature extraction module is used for extracting features of the image data in the target data set and the training data set respectively to obtain a target data feature vector and a training data feature vector; the feature fusion module is used for carrying out distinguishing feature fusion on the target data feature vector and the training data feature vector to obtain a fusion feature vector; The image generation module is used for generating fusion image data corresponding to the fusion feature vector according to the fusion feature vector; And the data updating module is used for updating the image data in the training data set into the fused image data. A third aspect of an embodiment of the present application provides a terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing a data enhancement method as described above when executing the computer program. A fourth aspect of the embodiments of the present application provides a computer-readable storage medium storing a computer program which, when executed by a processor, implements a data enhancement method as described above. According to the data enhancement method provided by the first aspect of the embodiment of the application, the fusion feature vector is obtained by carrying out the distinguishing feature fusion on the target data feature vector extracted by the target data set and the training data feature vector extracted by the training data set, the corresponding fusion image data is generated according to the fusion feature vector, the image data in the training data set is updated into the fusion image data, the data enhancement direction of the training data set is not random, and when individual difference data appear in the image data in the target data set, the distinguishing features of the individual difference data can be fused into the image data in the training data set, so that the image data in the enhanced training data set can also cover the individual difference data. It will be appreciated that the advantages of the second, third and fourth aspects may be found in the relevant description of the first aspect and are not repeated here. Drawings In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present application, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art. Fig. 1 is a schematic flow chart of a data enhancement method according to an embodiment of the present application; FIG. 2 is a schematic diagram of a second flow chart of a data enhancement method according to an embodiment of the present application; FIG. 3 is