A Data Enhancement Method, System, and Storage Medium in a Multi-Data Mode Environment

Abstract

The present invention provides a data enhancement method, system and storage medium in a multi-data mode environment. The method includes: confirming a specified input folder that stores original sample data for training; after the confirmation of the input folder is completed, creating a specified output folder; selecting a corresponding system operation mode and data enhancement algorithm according to the data modality required by the user; performing data enhancement processing on the original sample data in the input folder according to the selected system operation mode and data enhancement algorithm, and generating corresponding enhanced sample data; and saving the original sample data and the enhanced sample data in the output folder. In view of the current scattered situation of data enhancement algorithms, the present invention can quickly complete various data enhancement processing operations on training samples by aggregating multiple system operation modes and corresponding data enhancement algorithms, thereby improving work efficiency.

Description

Technical Field

[0001] The present invention relates to the field of computer vision technology, and particularly to a data augmentation method, system and storage medium in a multi-data mode environment. Background Art

[0002] In the field of deep learning in artificial intelligence, various network models obtain information and patterns from a vast amount of data. And in order to avoid the problem of overfitting, researchers usually need to input a sufficient amount of data. However, in many application scenarios, there are not enough training samples, such as medical image analysis, object detection in remote sensing satellite images, etc. Therefore, data augmentation is an essential part in the field of deep learning.

[0003] Data augmentation, also known as data amplification, aims to make limited data generate the value equivalent to more data without substantially increasing the data. Specifically, data augmentation can be divided into: supervised data augmentation and unsupervised data augmentation methods. Among them, supervised data augmentation can be further divided into single-sample data augmentation and multi-sample data augmentation methods, and unsupervised data augmentation is divided into two directions: generating new data and learning augmentation strategies.

[0004] Generally, data amplification refers to single-sample data augmentation of supervised data augmentation. Briefly speaking, supervised data augmentation is to complete the amplification of data by using preset data transformation rules on the basis of existing data. Single-sample data augmentation, that is, when enhancing a sample, all operations are carried out around the sample itself, including geometric transformation types, color transformation types, etc. Geometric transformation types are to perform geometric transformations on images, including operations such as flipping, rotating, cropping, deforming, scaling, etc. Such operations do not change the content of the image itself, and may be to select a part of the image or redistribute pixels. If you want to change the content of the image itself, you can choose data augmentation of the color transformation type of the image, and common ones include noise, blur, color transformation, erasure, filling, etc.

[0005] Currently, data augmentation algorithms are seriously scattered, and there are few methods that can be used for the processing of multiple data modes. When performing data amplification of multiple algorithms on training samples, several script files are required to implement. In addition, when processing training samples in different research fields, more algorithms are needed, the operation is more complex, and the time consumption is also more. Summary of the Invention

[0006] In order to solve the above at least one technical problem, the present invention proposes a data augmentation method, system and storage medium in a multi-data mode environment.

[0007] To achieve the above object, the first aspect of the present invention proposes a data augmentation method in a multi-data mode environment, and the method includes:

[0008] Confirm the specified input folder, which stores the original sample data for training;

[0009] After the confirmation of the input folder is completed, create the specified output folder;

[0010] Select the corresponding system operation mode and data augmentation algorithm according to the data modality required to be processed by the user;

[0011] Perform data augmentation processing on the original sample data in the input folder according to the selected system operation mode and data augmentation algorithm, and generate the corresponding augmented sample data;

[0012] Save the original sample data and the augmented sample data in the output folder.

[0013] In this solution, after creating the specified output folder, the method further includes:

[0014] Select the object detection mode according to the data modality required to be processed by the user;

[0015] Select the corresponding data augmentation algorithm in the object detection mode;

[0016] Perform augmentation processing on the original sample data including the original image and its corresponding XML file according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding XML file.

[0017] In this solution, after creating the specified output folder, the method further includes:

[0018] Select the semantic segmentation mode according to the data modality required to be processed by the user;

[0019] Select the corresponding data augmentation algorithm in the semantic segmentation mode;

[0020] Perform augmentation processing on the original sample data including the original image and its corresponding label image according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding label image.

[0021] In this solution, before selecting the corresponding system operation mode and data augmentation algorithm according to the data modality required to be processed by the user, the method further includes:

[0022] Receive the user's request to add a new system operation mode;

[0023] Add a new system operation mode according to the request.

[0024] In this solution, before selecting the corresponding system operation mode and data augmentation algorithm according to the data modality required by the user, the method further includes:

[0025] Receiving a request from the user to add a new data augmentation algorithm for a specified system operation mode;

[0026] Adding the new data augmentation algorithm under the specified system operation mode according to the request.

[0027] In this solution, after saving the original sample data and the augmented sample data in the output folder, the method further includes:

[0028] Performing model training based on the original sample data and the augmented sample data in the output folder.

[0029] Preferably, the data augmentation algorithm is any one or several of a filtering algorithm, a flipping algorithm, a rotation algorithm, a noise algorithm, a chromaticity enhancement algorithm, a scaling algorithm, and a cropping algorithm.

[0030] The second aspect of the present invention further proposes a data augmentation system in a multi-data mode environment. The data augmentation system in the multi-data mode environment includes: a memory and a processor. The memory includes a data augmentation method program in a multi-data mode environment. When the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are implemented:

[0031] Confirming a specified input folder that stores the original sample data for training;

[0032] After the confirmation of the input folder is completed, creating a specified output folder;

[0033] Selecting a corresponding system operation mode and data augmentation algorithm according to the data modality required by the user;

[0034] Performing data augmentation processing on the original sample data in the input folder according to the selected system operation mode and data augmentation algorithm, and generating corresponding augmented sample data;

[0035] Saving the original sample data and the augmented sample data in the output folder.

[0036] In this solution, when the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0037] Selecting a target detection mode or a semantic segmentation mode according to the data modality required by the user;

[0038] In the target detection mode, select the corresponding data augmentation algorithm, and perform augmentation processing on the original sample data including the original image and its corresponding XML file according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding XML file; in the semantic segmentation mode, select the corresponding data augmentation algorithm, and perform augmentation processing on the original sample data including the original image and its corresponding label image according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding label image.

[0039] The third aspect of the present invention also proposes a computer-readable storage medium, which includes a data augmentation method program in a multi-data mode environment. When the data augmentation method program in the multi-data mode environment is executed by a processor, the steps of a data augmentation method in a multi-data mode environment as described above are implemented.

[0040] The present invention discloses a data augmentation method, system and storage medium in a multi-data mode environment, which can be used for data amplification of training samples in fields such as target detection and semantic segmentation, and can expand the system operation mode and data augmentation algorithm. Users can utilize this system framework to quickly complete the enhancement of the size and quality of the training sample dataset through simple and convenient operations, so as to complete deep learning tasks with higher quality.

[0041] In view of the current situation that data augmentation algorithms are relatively scattered, the present invention can simply and conveniently quickly complete various data augmentation processing operations on training samples by aggregating multiple system operation modes and corresponding data augmentation algorithms, thereby improving work efficiency. In addition, the present invention can play a positive role in promoting the research and application development in the field of artificial intelligence.

[0042] The additional aspects and advantages of the present invention will be given in the following description section, some will become obvious from the following description, or will be understood through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 Shows a flowchart of a data augmentation method in a multi-data mode environment of the present invention;

[0044] Figure 2 Shows a flowchart of a data augmentation method in the target detection mode of the present invention;

[0045] Figure 3 Shows a flowchart of a data augmentation method in the semantic segmentation mode of the present invention;

[0046] Figure 4 Shows a block diagram of a data augmentation system in a multi-data mode environment of the present invention;

[0047] Figure 5 Shows the operation flowchart of the data enhancement system in the multi - data mode environment of the present invention;

[0048] Figure 6 Shows the effect diagram of the data enhancement example in the semantic segmentation mode of the present invention. Detailed implementation manners

[0049] In order to more clearly understand the above - mentioned objects, features and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0050] In the following description, many specific details are set forth to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0051] Figure 1 Shows the flowchart of a data enhancement method in a multi - data mode environment of the present invention.

[0052] As Figure 1 shown, a data enhancement method in a multi - data mode environment is proposed in the first aspect of the present invention, and the method includes:

[0053] S102, confirm a specified input folder, where the input folder stores the original sample data for training;

[0054] S104, after the confirmation of the input folder is completed, create a specified output folder;

[0055] S106, select the corresponding system operation mode and data enhancement algorithm according to the data modality that the user needs to process;

[0056] S108, perform data enhancement processing on the original sample data in the input folder according to the selected system operation mode and data enhancement algorithm, and generate corresponding enhanced sample data;

[0057] S110, save the original sample data and the enhanced sample data in the output folder.

[0058] It should be noted that the technical solution of the present invention can be implemented in terminal devices such as PCs, mobile phones, and PADs.

[0059] It should be noted that the original sample data can be saved in the form of folders. Since different folders may store different types of original sample data. When training a certain model, it is first necessary to determine the original sample data suitable for training the model and the folder where it is located, and use this folder as the input folder.

[0060] It should be noted that the system operation mode can include any one or several of the object detection mode, semantic segmentation mode, and classification and localization mode, but is not limited thereto.

[0061] It should be noted that the original sample data can include the original image and its corresponding original label data; the enhanced sample data can include the enhanced image and its corresponding enhanced label data. It can be understood that the enhanced image is obtained by enhancing the corresponding original image, and the enhanced label data is obtained by enhancing the corresponding original label data.

[0062] It should be noted that the data enhancement algorithm can be any one or several of the filtering algorithm, flipping algorithm, rotation algorithm, noise algorithm, chromaticity enhancement algorithm, scaling algorithm, and cropping algorithm. But is not limited thereto.

[0063] It can be understood that in a certain system operation mode, when multiple data enhancement algorithms are selected, the original sample data can be processed by multiple data enhancement algorithms at the same time. Each data enhancement algorithm can double the amount of original sample data, and multiple data enhancement algorithms can multiply the amount of original sample data. Therefore, the present invention can achieve the amplification of the sample data volume without increasing the original sample data, effectively avoiding the phenomenon of overfitting in model training caused by insufficient original sample data volume.

[0064] It should be noted that after the sample data amplification is completed, the present invention can train the model based on the original sample data and the enhanced sample data in a supervised learning manner. Specifically, the original image and its corresponding original label data, and the enhanced image and its corresponding enhanced label data can be input into the model for analysis respectively. After the model completes learning, when an unknown image is input, the predicted label of the image can be output. Since the model not only has training data (such as the original image and the enhanced image) but also has training results (such as the original label data and the enhanced label data) as optimization support during the learning process, the training effect of the model is better.

[0065] Figure 2 The flowchart of the data enhancement method in the object detection mode of the present invention is shown.

[0066] As Figure 2 shown, after creating the specified output folder, the method further includes:

[0067] S202. Select a target detection mode according to the data modality required to be processed by the user.

[0068] S204. In the target detection mode, select the corresponding data augmentation algorithm.

[0069] S206. Perform augmentation processing on the original sample data including the original image and its corresponding XML file according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding XML file.

[0070] It should be noted that the XML file includes the label information corresponding to the original image. When the user needs to process the training sample data in the field of target detection, the target detection mode can be selected. At this time, the data augmentation algorithm can perform data augmentation processing on both the original image and the XML file containing the label information. Preferably, the data augmentation algorithm in the target detection mode can include any one or several of the flipping algorithm, rotation algorithm, noise perturbation algorithm, scaling algorithm, and cropping algorithm. But it is not limited to this.

[0071] It can be understood that so-called target detection means inputting an image into the model, and whenever a certain type of object appears in the image, a box is drawn around the object and the category to which the object belongs is predicted.

[0072] Figure 3 The flowchart of the data augmentation method in the semantic segmentation mode of the present invention is shown.

[0073] As Figure 3 shown, after creating the specified output folder, the method further includes:

[0074] S302. Select a semantic segmentation mode according to the data modality required to be processed by the user.

[0075] S304. In the semantic segmentation mode, select the corresponding data augmentation algorithm.

[0076] S306. Perform augmentation processing on the original sample data including the original image and its corresponding label image according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding label image.

[0077] It should be noted that the label image can be a label mask image. When the user needs to process the training sample data in the field of semantic segmentation, the semantic segmentation mode can be selected. At this time, the data augmentation algorithm can perform data augmentation processing on the original image and its corresponding label mask image simultaneously. Preferably, the data augmentation algorithm in the semantic segmentation mode can include any one or several of a filtering algorithm, a flipping algorithm, a rotation algorithm, a noise algorithm, and a chromaticity enhancement algorithm. However, it is not limited thereto.

[0078] It can be understood that so-called semantic segmentation means inputting an image into a model and classifying each pixel in the image.

[0079] According to an embodiment of the present invention, before selecting the corresponding system operation mode and data augmentation algorithm according to the data modality required by the user, the method further includes:

[0080] Receiving a request from the user to add a new system operation mode;

[0081] Adding a new system operation mode according to the request.

[0082] It should be noted that the present invention can freely add a new system operation mode in the system framework according to the user's needs.

[0083] Furthermore, the addition permissions for different user identities can also be managed. When the system receives a request from the user to add a new system operation mode, the system automatically identifies the user's identity and determines the addition permissions of this user identity. If the new system operation mode is within the addition permissions of this user, adding this system operation mode in the system framework is allowed; if the new system operation mode exceeds the addition permissions of this user, adding this system operation mode in the system framework is rejected. At this time, if it is necessary to continue adding this system operation mode, an application can be made to the administrator to add this system operation mode, or the user's permission to add this system operation mode can be granted.

[0084] According to an embodiment of the present invention, the method further includes:

[0085] Receiving a request from the user to delete the corresponding system operation mode;

[0086] Deleting this system operation mode in the system framework according to the request.

[0087] It should be noted that the present invention can freely delete the corresponding system operation mode in the system framework according to the user's needs.

[0088] Further, it is also possible to manage the deletion permissions for different user identities. When the system receives a request from a user to delete a corresponding system operation mode, the system automatically identifies the user's identity and determines the deletion permissions for that user identity. If the system operation mode to be deleted is within the deletion permissions of that user, then permission is granted to delete the system operation mode in the system framework; if the system operation mode to be deleted exceeds the deletion permissions of that user, then the deletion of the system operation mode in the system framework is refused. At this time, if it is necessary to continue deleting the system operation mode, then an application can be made to the administrator to delete the system operation mode, or to grant the user the permission to delete the system operation mode.

[0089] According to an embodiment of the present invention, the method further includes:

[0090] Receiving a request from a user to change a first system operation mode to a second system operation mode;

[0091] Changing the first system operation mode to the second system operation mode in the system framework according to the request.

[0092] It should be noted that the present invention can freely change the corresponding system operation mode according to the user's needs in the system framework.

[0093] Further, it is also possible to manage the change permissions for different user identities. When the system receives a request from a user to change a corresponding system operation mode, the system automatically identifies the user's identity and determines the change permissions for that user identity. If the system operation mode to be changed is within the change permissions of that user, then permission is granted to change the system operation mode in the system framework; if the system operation mode to be changed exceeds the change permissions of that user, then the change of the system operation mode in the system framework is refused. At this time, if it is necessary to continue changing the system operation mode, then an application can be made to the administrator to change the system operation mode, or to grant the user the permission to change the system operation mode.

[0094] According to an embodiment of the present invention, before selecting a corresponding system operation mode and a data enhancement algorithm according to the data modality to be processed by the user, the method further includes:

[0095] Receiving a request from a user to add a new data enhancement algorithm to a specified system operation mode;

[0096] Adding the new data enhancement algorithm under the specified system operation mode according to the request.

[0097] It should be noted that the present invention can freely add new data enhancement algorithms under each system operation mode.

[0098] Further, it is also possible to manage the addition permissions for different user identities in each system operation mode. When the system receives a request from a user to add a new data enhancement algorithm for a specified system operation mode, the system automatically identifies the user's identity and determines the addition permissions for this user identity. If the specified system operation mode is within the addition permissions of this user, then adding a new data enhancement algorithm is allowed in this system operation mode; if the specified system operation mode exceeds the addition permissions of this user, then adding a new data enhancement algorithm in this system operation mode is rejected. At this time, if it is necessary to continue adding a new data enhancement algorithm in this system operation mode, an application can be made to the administrator to add a new data enhancement algorithm, or to grant this user the permission to add a data enhancement algorithm in this system operation mode.

[0099] According to an embodiment of the present invention, the method further includes:

[0100] Receiving a request from a user to delete a corresponding data enhancement algorithm for a specified system operation mode;

[0101] Deleting the data enhancement algorithm in the specified system operation mode according to the request.

[0102] According to an embodiment of the present invention, the method further includes:

[0103] Receiving a request from a user to replace a first data enhancement algorithm with a second data enhancement algorithm for a specified system operation mode;

[0104] Replacing the first data enhancement algorithm with the second data enhancement algorithm in the specified system operation mode according to the request.

[0105] It should be noted that the present invention can freely delete or replace data enhancement algorithms in each system operation mode.

[0106] Further, it is also possible to manage the deletion or replacement permissions for different user identities in each system operation mode. When the system receives a request from a user to delete or replace a corresponding data enhancement algorithm for a specified system operation mode, the system automatically identifies the user's identity and determines the deletion or replacement permissions for this user identity. If the specified system operation mode is within the deletion or replacement permissions of this user, then deleting or replacing the data enhancement algorithm is allowed in this system operation mode; if the specified system operation mode exceeds the deletion or replacement permissions of this user, then deleting or replacing the data enhancement algorithm in this system operation mode is rejected. At this time, if it is necessary to continue deleting or replacing the data enhancement algorithm in this system operation mode, an application can be made to the administrator to delete or replace the data enhancement algorithm, or to grant this user the permission to delete or replace the data enhancement algorithm in this system operation mode.

[0107] After saving the original sample data and the augmented sample data in the output folder according to an embodiment of the present invention, the method further includes:

[0108] Perform model training based on the original sample data and the augmented sample data in the output folder.

[0109] By integrating processing algorithms of multiple data modalities, the above method can freely and conveniently perform several data augmentation operations on the training sample data, providing more sufficient training samples for the model and effectively avoiding the phenomenon of overfitting.

[0110] Figure 4 The block diagram of a data augmentation system in a multi-data mode environment according to the present invention is shown.

[0111] As Figure 4 shown, in a second aspect of the present invention, a data augmentation system 4 in a multi-data mode environment is further proposed. The data augmentation system 4 in the multi-data mode environment includes: a memory 41 and a processor 42. The memory includes a data augmentation method program in a multi-data mode environment. When the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are implemented:

[0112] Confirm a specified input folder that stores the original sample data for training;

[0113] After the confirmation of the input folder is completed, create a specified output folder;

[0114] Select a corresponding system operation mode and data augmentation algorithm according to the data modality required to be processed by the user;

[0115] Perform data augmentation processing on the original sample data in the input folder according to the selected system operation mode and data augmentation algorithm, and generate corresponding augmented sample data;

[0116] Save the original sample data and the augmented sample data in the output folder.

[0117] It should be noted that the system of the present invention can be operated on terminal devices such as PCs, mobile phones, and PADs.

[0118] It should be noted that the processor may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0119] It should be noted that the data augmentation algorithm may be any one or several of a filtering algorithm, a flipping algorithm, a rotation algorithm, a noise algorithm, a chromaticity enhancement algorithm, a scaling algorithm, and a cropping algorithm. However, it is not limited thereto.

[0120] According to an embodiment of the present invention, when the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0121] Select a target detection mode according to the data modality required to be processed by the user;

[0122] In the target detection mode, select the corresponding data augmentation algorithm;

[0123] Perform augmentation processing on the original sample data including the original image and its corresponding XML file according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding XML file.

[0124] According to an embodiment of the present invention, when the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0125] Select a semantic segmentation mode according to the data modality required to be processed by the user;

[0126] In the semantic segmentation mode, select the corresponding data augmentation algorithm;

[0127] Perform augmentation processing on the original sample data including the original image and its corresponding label image according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding label image.

[0128] According to an embodiment of the present invention, when the data augmentation method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0129] Receive a request from the user to add a new system operation mode;

[0130] Add a new system operation mode according to the said request.

[0131] According to an embodiment of the present invention, when the data enhancement method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0132] Receive a request from the user to add a new data enhancement algorithm for a specified system operation mode;

[0133] Add the said new data enhancement algorithm under the specified system operation mode according to the request.

[0134] According to an embodiment of the present invention, when the data enhancement method program in the multi-data mode environment is executed by the processor, the following steps are further implemented:

[0135] Perform model training based on the original sample data and the enhanced sample data in the output folder.

[0136] A third aspect of the present invention further proposes a computer-readable storage medium, which includes a data enhancement method program in a multi-data mode environment. When the data enhancement method program in the multi-data mode environment is executed by a processor, the steps of a data enhancement method in a multi-data mode environment as described above are implemented.

[0137] To further explain the technical solution of the present invention, a detailed description will be given below with an embodiment.

[0138] As Figure 5 shown, the specific operation process of the data enhancement system in this embodiment is as follows: First, confirm the folder of the data to be processed and use this folder as the input folder of the system; after the input folder is confirmed, create an output folder; then select the system operation mode. In this embodiment, the system operation mode includes an object detection mode and a semantic segmentation mode, and the user can select between these two modes; when the semantic segmentation mode is selected, the data enhancement algorithm in this semantic segmentation mode performs enhancement processing on the original sample data in the input folder and generates enhanced sample data. As Figure 6 shown, the data enhancement algorithm includes a rotation algorithm, a flipping algorithm, and a Gaussian noise algorithm; the original sample data includes the original image and its corresponding label; the enhanced sample data includes the rotated image and its corresponding label, the flipped image and its corresponding label, and the image after Gaussian noise processing and its corresponding label.

[0139] The present invention discloses a data enhancement method, system and storage medium in a multi-data mode environment, which can be used for data augmentation of training samples in fields such as object detection and semantic segmentation, and can also expand the system operation mode and data enhancement algorithm. Users can utilize this system framework to quickly complete the enhancement of the size and quality of the training sample dataset through simple and convenient operations, thereby more efficiently completing deep learning tasks.

[0140] In view of the current scattered situation of data enhancement algorithms, the present invention can simply and conveniently complete various data enhancement processing operations on training samples by adding various system operation modes and corresponding data enhancement algorithms, thereby improving work efficiency. In addition, the present invention can play a positive role in promoting the research and application development in the field of artificial intelligence.

[0141] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical, or other forms.

[0142] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units; they can be located in one place or distributed to multiple network units; some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0143] In addition, in each embodiment of the present invention, the various functional units can all be integrated in one processing unit, or each unit can be separately used as a unit, or two or more units can be integrated in one unit; the above integrated units can be implemented in the form of hardware, or in the form of a combination of hardware and software functional units.

[0144] Those of ordinary skill in the art will understand that all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions. The foregoing program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps including those of the above method embodiments; and the foregoing storage medium includes: removable storage devices, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks and other various media that can store program codes.

[0145] Alternatively, if the above integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in the various embodiments of the present invention. And the foregoing storage medium includes: removable storage devices, ROM, RAM, magnetic disks, or optical disks and other various media that can store program codes.

[0146] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A data augmentation method in a multi - data - mode environment, characterized in that, the method includes: Confirm a specified input folder that stores the original sample data for training; After the confirmation of the input folder is completed, create a specified output folder; Select a corresponding system operation mode and a data augmentation algorithm according to the data modality required to be processed by the user; wherein, the system operation mode includes at least one of the following: object detection mode, semantic segmentation mode, classification and localization mode; the data augmentation algorithm is any one or several of a filtering algorithm, a flipping algorithm, a rotation algorithm, a noise algorithm, a chromaticity enhancement algorithm, a scaling algorithm, and a cropping algorithm; Perform data augmentation processing on the original sample data in the input folder according to the selected system operation mode and data augmentation algorithm, and generate corresponding augmented sample data; Save the original sample data and the augmented sample data in the output folder; wherein, After creating the specified output folder, the method includes: Select the semantic segmentation mode according to the data modality required to be processed by the user; Under the semantic segmentation mode, select the corresponding data augmentation algorithm; Perform augmentation processing on the original sample data including the original image and its corresponding label image according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding label image.

2. The data augmentation method in a multi - data - mode environment according to claim 1, characterized in that, After creating the specified output folder, the method further includes: Select the object detection mode according to the data modality required to be processed by the user; Under the object detection mode, select the corresponding data augmentation algorithm; Perform augmentation processing on the original sample data including the original image and its corresponding XML file according to the selected data augmentation algorithm, and generate augmented sample data including the augmented image and its corresponding XML file.

3. The data augmentation method in a multi - data - mode environment according to claim 1, characterized in that, Before selecting the corresponding system operation mode and data augmentation algorithm according to the data modality required to be processed by the user, the method further includes: Receive a request from the user to add a new system operation mode; Add a new system operation mode according to the request.

4. The data augmentation method in a multi - data - mode environment according to claim 1, characterized in that, Before selecting the corresponding system operation mode and data augmentation algorithm according to the data modality required to be processed by the user, the method further includes: Receive a request from the user to add a new data augmentation algorithm for a specified system operation mode; Add the new data augmentation algorithm under the specified system operation mode according to the request.

5. The data augmentation method in a multi - data - mode environment according to claim 1, characterized in that, After saving the original sample data and the augmented sample data in the output folder, the method further includes: Perform model training based on the original sample data and the augmented sample data in the output folder.

6. A data enhancement system in a multi-data mode environment, characterized in that, the data enhancement system in the multi-data mode environment includes: a memory and a processor, and the memory includes a data enhancement method program in the multi-data mode environment. When the data enhancement method program in the multi-data mode environment is executed by the processor, the following steps are implemented: Confirm a specified input folder, where the input folder stores the original sample data for training; After the confirmation of the input folder is completed, create a specified output folder; Select the corresponding system operation mode and data enhancement algorithm according to the data modality to be processed by the user; wherein, the system operation mode includes at least one of the following: object detection mode, semantic segmentation mode, classification and localization mode; the data enhancement algorithm is any one or several of a filtering algorithm, a flipping algorithm, a rotation algorithm, a noise algorithm, a chromaticity enhancement algorithm, a scaling algorithm, and a cropping algorithm; Perform data enhancement processing on the original sample data in the input folder according to the selected system operation mode and data enhancement algorithm, and generate corresponding enhanced sample data; Save the original sample data and the enhanced sample data in the output folder; wherein, after creating the specified output folder, the method further includes: Select the semantic segmentation mode according to the data modality to be processed by the user; In the semantic segmentation mode, select the corresponding data enhancement algorithm; Perform enhancement processing on the original sample data including the original image and its corresponding label image according to the selected data enhancement algorithm, and generate enhanced sample data including the enhanced image and its corresponding label image.

7. The data enhancement system in a multi-data mode environment according to claim 6, characterized in that, when the data enhancement method program in the multi-data mode environment is executed by the processor, the following steps are further implemented: Select the object detection mode according to the data modality to be processed by the user; In the object detection mode, select the corresponding data enhancement algorithm, and perform enhancement processing on the original sample data including the original image and its corresponding XML file according to the selected data enhancement algorithm, and generate enhanced sample data including the enhanced image and its corresponding XML file.

8. A computer-readable storage medium, characterized in that, the computer-readable storage medium includes a data enhancement method program in a multi-data mode environment. When the data enhancement method program in the multi-data mode environment is executed by a processor, the steps of a data enhancement method in a multi-data mode environment as described in any one of claims 1 to 5 are implemented.

Images