Method and device for expanding training data set through computing power of intelligent computing center

Abstract

The invention provides a method and device for expanding a training data set through computing power of an intelligent computing center, and relates to the technical field of intelligent computing centers, intelligent computing centers, computing power infrastructures and automobiles, and the method comprises the steps: obtaining a first training data set, an abnormal identifier inputted by a user, and size information corresponding to the abnormal identifier; identifying the first picture according to the size information to obtain a blank area of the first picture; generating a second picture based on the blank area and the abnormal identifier; and in the first training data set, replacing the first picture with the second picture to obtain a second training data set. Through the above steps, the second picture which contains the abnormal identifier and is different from the first picture can be continuously generated, and different types of abnormal data can be automatically amplified in different detection scenes until the data volume of the second training data set can support the training of the image recognition model so as to improve the training effect of the model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent computing centers, wisdom computing centers, computing power infrastructures and automobiles, in particular to a method and device for expanding a training data set by computing power of an intelligent computing center. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "wisdom computing centers" have emerged.

[0003] An "intelligent computing center" refers to a facility that provides required computing power, data and algorithms for artificial intelligence applications (such as artificial intelligence deep learning model development, model training and model inference scenarios) by using large-scale heterogeneous computing power resources, including general computing power and intelligent computing power. An intelligent computing center covers facilities, hardware and software, and can provide full-stack capabilities from bottom computing power to top application enablement.

[0004] An "intelligent computing center" includes but is not limited to a "wisdom computing center".

[0005] A "wisdom computing center" is an artificial intelligence computing center based on artificial intelligence theory, using artificial intelligence computing architecture to provide computing power services, data services and algorithm services required by artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "wisdom computing centers". It is the ability of computer equipment or computing / data centers to process information. It is the ability of computer hardware and software to jointly execute certain computing needs. It is the computing power that achieves target result output by processing information data. It is a new type of productivity that integrates information computing power, network carrying capacity and data storage power, and mainly provides services to society through computing power infrastructure.

[0007] Currently, in the process of model training by computing power of an intelligent computing center, especially for models used to identify whether a picture is abnormal (for example, identifying tumors in medical image pictures or identifying abnormal marks in two-dimensional drawings in the automotive field), the number of abnormal data is much lower than the number of normal data, which results in a lack of data in the training data set collected for model training, which cannot support model training, and leads to poor model recognition ability for picture abnormalities, i.e. poor model training effect. SUMMARY

[0008] The present application provides a method and device for expanding a training data set by computing power of an intelligent computing center, which is used to solve the problem of poor model training effect.

[0009] In order to solve the above technical problems, the present application is implemented as follows:

[0010] In a first aspect, the present application provides a method for expanding a training data set by computing power of an intelligent computing center, comprising:

[0011] Step S1: obtaining a first training data set, a user input abnormality identifier, and size information corresponding to the abnormality identifier, the first training data set comprising a first picture;

[0012] Step S2: identifying the first picture according to the size information to obtain a blank area of the first picture, the size of the blank area being greater than or equal to the size indicated by the size information;

[0013] Step S3: generating a second picture based on the blank area and the abnormality identifier;

[0014] Step S4: replacing the first picture with the second picture in the first training data set to obtain a second training data set, the second training data set being used for training an image recognition model.

[0015] Optionally, the step S2 comprises:

[0016] Step S21: obtaining a random number N, N being a natural number;

[0017] Step S22: determining N blank areas from the first picture based on the size information.

[0018] Optionally, the step S22 comprises:

[0019] Step S221: obtaining a first center point;

[0020] Step S222: determining a first area corresponding to the first center point based on the first center point and the size information, and obtaining color information of the first area, the size of the first area being greater than or equal to the size indicated by the size information, the first area being formed by expanding the corresponding first center point to the surrounding;

[0021] Step S223: in the case where the color information of the first area indicates that the corresponding area is blank, determining the first area as the blank area;

[0022] Step S224: repeating steps S221 to S223 until a target condition occurs, the target condition comprising one of:

[0023] the number of the determined blank areas is N;

[0024] the number of the determined blank areas is less than N, and no more blank areas can be determined in the first picture.

[0025] Optionally, the step S22 comprises:

[0026] Step S225: obtaining M different second center points in the first picture, M being an integer greater than or equal to N;

[0027] Step S226: obtaining chrominance information of a first region corresponding to each of the second center points, the first region having a size greater than or equal to the size indicated by the size information, the first region being formed by expanding the corresponding second center point in all directions;

[0028] Step S227: determining N first regions from the M first regions as the blank regions, the N first regions having chrominance information indicating that the corresponding regions are blank.

[0029] Optionally, the step S3 comprises:

[0030] Step S31: obtaining a center point of the abnormality identifier;

[0031] Step S32: adding the N abnormality identifiers to the N blank regions one by one according to the center point of the abnormality identifier and the N center points of the N blank regions, to generate the second picture, the abnormality identifier being contained in the corresponding blank region in the second picture.

[0032] Optionally, after the step S4, the method further comprises:

[0033] Step S5: cyclically performing the steps S1-S4 to obtain a training data set containing multiple abnormality identifiers, the training data set containing multiple abnormality identifiers being used to train an image recognition model.

[0034] In a second aspect, the present application provides an apparatus for expanding a training data set, the apparatus comprising:

[0035] an obtaining module configured to obtain a first training data set, an abnormality identifier input by a user, and size information corresponding to the abnormality identifier, the first training data set comprising a first picture;

[0036] an identifying module configured to identify the first picture according to the size information to obtain a blank region of the first picture, the blank region having a size greater than or equal to the size indicated by the size information;

[0037] a generating module configured to generate a second picture based on the blank region and the abnormality identifier;

[0038] An adding module is configured to replace the first picture with the second picture in the first training data set to obtain a second training data set, and the second training data set is used to train the image recognition model.

[0039] In a third aspect, the present application provides an electronic device, comprising a processor, a memory, and a program stored in the memory and executable on the processor, and the program, when executed by the processor, implements the steps of the method for training data set by computing power expansion of an intelligent computing center according to the first aspect.

[0040] In a fourth aspect, the present application provides a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program, when executed by a processor, implements the steps of the method for training data set by computing power expansion of an intelligent computing center according to the first aspect.

[0041] In a fifth aspect, the present application provides a computer program product, comprising computer instructions, and the computer instructions, when executed by a processor, implement the steps of the method for training data set by computing power expansion of an intelligent computing center according to the first aspect.

[0042] In the present application, the first training data set, the user input abnormality identifier, and the size information corresponding to the abnormality identifier are obtained; the first picture is identified according to the size information to obtain the blank area of the first picture; the second picture is generated based on the blank area and the abnormality identifier; and the first picture is replaced with the second picture in the first training data set to obtain a second training data set. Through the above steps, the second picture containing the abnormality identifier and different from the first picture can be continuously generated until the data amount of the second training data set can support the training of the image recognition model, so as to improve the training effect of the model. BRIEF DESCRIPTION OF DRAWINGS

[0043] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of preferred embodiments, and are not meant to limit the present application. Moreover, the same reference numerals in the attached drawings refer to the same or similar components. In the drawings:

[0044] Figure 1 A flowchart of the method for training data set by computing power expansion of an intelligent computing center provided by the embodiments of the present application is shown;

[0045] Figure 2 A schematic diagram of the first picture provided by the embodiments of the present application is shown;

[0046] Figure 3A structural schematic diagram of the device for expanding the training data set provided by the embodiment of the present application is shown in the figure.

[0047] Figure 4 A structural schematic diagram of the electronic device provided by the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0048] The technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0049] First, the technical terms related to the present application will be described simply.

[0050] The "computing power" in the present application refers to the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to jointly perform certain computing requirements, the computing power to achieve target result output by processing information data, and a new type of productivity integrating information computing power, network carrying capacity and data storage power, which mainly provides services to the society through computing power infrastructure.

[0051] The "computing power" (Computational Power, CP) in the present application refers to the ability of a data center server to process data and achieve result output, which is a comprehensive index for measuring the computing power of a data center, including general computing power, supercomputing power and intelligent computing power. The commonly used unit of measurement is the number of floating point operations per second (FLOPS, 1EFLOPS=10^18FLOPS), and the larger the value is, the stronger the comprehensive computing power is. According to the estimation, 1EFLOPS is about the computing power output of 5 Tianhe 2A, 500,000 mainstream server CPUs or 2,000,000 mainstream notebooks. The calculation formula is: CP=CP 通用 +CP 智能 +CP 超级 .

[0052] The "carrying capacity" (Network Power, NP) in the present application refers to the performance of the data transmission capacity of computing power facilities, including the comprehensive ability of network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., involving the network transmission between the data centers, and being a comprehensive index for measuring the network transmission scheduling ability.

[0053] The "storage power" (SP) described in the application refers to the comprehensive ability of a data center in four aspects of data storage capacity, performance, safety and reliability, and green and low carbon, which is a comprehensive index for measuring the data storage capacity of a data center, and includes external storage devices such as storage arrays and built-in storage devices of servers. The commonly used unit of measurement for storage capacity is exabyte (EB, 1 EB = 2^60 bytes), the commonly used unit of measurement for performance is the number of read and write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB), and the disaster recovery ratio is an important performance of safety and reliability.

[0054] The "computing power infrastructure" described in the application refers to a new type of information infrastructure integrating information computing power, network carrying capacity and data storage power, which can realize centralized computing, storage, transmission and application of information.

[0055] The "new type of information infrastructure" described in the application refers to network infrastructure including 5G network, fiber broadband network, backbone network, international communication network, satellite Internet, etc., computing power infrastructure including data center, general computing center, intelligent computing center, supercomputing center, etc., and new technology infrastructure including artificial intelligence, blockchain, quantum computing, etc.

[0056] The "computing power" described in the application includes general computing power, intelligent computing power and supercomputing power.

[0057] The "general computing power" described in the application refers to the computing power provided by a server based on a CPU (Central Processing Unit) chip, which is used to support basic general computing such as cloud computing and edge computing.

[0058] The "intelligent computing power" described in the application refers to the large-scale deployment of computing platforms based on GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array) and ASIC (Application Specific Integrated Circuit) special chips for various artificial intelligence innovation applications, such as natural language processing and machine vision.

[0059] The "supercomputing power" described in the application refers to the computing power provided by a supercomputer and other high-performance computing clusters, which utilizes the centralized computing resources of multiple computer systems working in parallel and processes extremely complex or data-intensive problems through a dedicated operating system, mainly used for computing in cutting-edge scientific fields such as planet simulation, drug molecule design and gene analysis.

[0060] The "intelligent computing center" described in this invention refers to a facility that, through the use of large-scale heterogeneous computing resources, including general-purpose computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.), primarily provides the necessary computing power, data, and algorithms for artificial intelligence applications (such as the development, training, and inference of deep learning models). The intelligent computing center encompasses facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0061] The "intelligent computing center" mentioned in this invention includes, but is not limited to, "smart computing center".

[0062] The "intelligent computing center" mentioned in this invention, also known as an artificial intelligence computing center, is a type of computing infrastructure that provides computing power services, data services, and algorithm services required for artificial intelligence applications, based on artificial intelligence theory and adopting an artificial intelligence computing architecture.

[0063] The "computing power center" mentioned in this invention refers to a facility that is mainly composed of infrastructure such as wind, thermal, and hydropower and IT hardware and software equipment, and has computing power, carrying capacity and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0064] The "supercomputing center" mentioned in this invention refers to a supercomputing data center, which is a data center based on supercomputers or large-scale computing clusters. It can provide large-scale computing, storage and network services and is widely used in aerospace, defense, oil exploration, climate modeling and genome sequencing and other application scenarios.

[0065] The “computing resources” mentioned in this invention refer to the technologies and facilities required for the development of the digital society that have the ability to compute, transmit, store and apply information, including but not limited to computing resources such as CPUs and GPUs, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and supporting and guaranteeing resources such as wind, fire, water and electricity.

[0066] like Figure 1 As shown, this embodiment of the invention provides a method for expanding a training dataset using the computing power of an intelligent computing center, comprising the following steps:

[0067] Step S1: Obtain the first training dataset, the anomaly identifier input by the user, and the size information corresponding to the anomaly identifier. The first training dataset includes the first image.

[0068] In this step, the first training data set is the data collected for training the image recognition model, including abnormal data and normal data, and the first picture can be abnormal data or normal data. The image recognition model is a model that automatically recognizes, classifies or detects objects, scenes, features or patterns in digital images or video frames by analyzing pixel data in the images or video frames.

[0069] With different specific application scenarios of the image recognition model, the collected abnormal data and normal data are also different. For example, if the application scenario is to identify tumors in medical imaging pictures, then the corresponding normal data is medical imaging pictures that do not contain tumors, and the abnormal data is medical imaging pictures that contain tumors. If the application scenario is to identify abnormal marks in two-dimensional drawings in the automotive field, then the corresponding normal data is two-dimensional drawings in the automotive field that do not contain abnormal marks, and the abnormal data is two-dimensional drawings in the automotive field that contain abnormal marks. It should be noted that in many application scenarios, there is a problem that the amount of abnormal data is too small to complete model training.

[0070] Step S2: identifying the first picture according to the size information to obtain a blank area of the first picture, the size of the blank area being greater than or equal to the size indicated by the size information.

[0071] As shown in Figure 2 The first picture includes a part of graphic content and a blank part without graphic content, and the blank area with a size greater than or equal to the size indicated by the size information can be determined from the blank part without graphic content.

[0072] Step S3: generating a second picture based on the blank area and the abnormal mark.

[0073] The size of the blank area is greater than or equal to the size of the abnormal mark, and the abnormal mark can be added to the blank area of the first picture to obtain a second picture containing the abnormal mark, that is, new abnormal data. It can be understood that repeating the above steps S2 and S3 can obtain several different second pictures containing abnormal marks, that is, several different abnormal data.

[0074] Step S4, in the first training data set, the first picture is replaced by the second picture, and the second training data set is used to train the image recognition model.

[0075] The second picture containing the abnormal mark and different from the abnormal data contained in the original first training data set is replaced by the first picture to obtain a second training data set, and repeating the above steps S2 to S4 can continuously increase the second picture, so that the amount of abnormal data in the second training data set can support the training of the image recognition model.

[0076] In the method for expanding a training data set through intelligent computing center computing power in the embodiment of the application, a first training data set, user input abnormality identification, and size information corresponding to the abnormality identification are obtained; the first picture is identified according to the size information to obtain a blank area of the first picture; a second picture is generated based on the blank area and the abnormality identification; and the first picture is replaced by the second picture in the first training data set to obtain a second training data set. Through the above steps, a second picture containing abnormality identification and different from the first picture can be continuously generated, different types of abnormal data can be automatically expanded in different detection scenarios, and the data amount of the second training data set can support the training of an image recognition model until the training effect of the model is improved.

[0077] The step S2 comprises:

[0078] Step S21: obtaining a random number N, N being a natural number;

[0079] Step S22: determining N blank areas from the first picture based on the size information.

[0080] In this embodiment, when N is 1, a blank area with a size greater than or equal to the size of the abnormality identification can be selected in the first picture. When N is greater than 1, blank areas need to be selected in the first picture one by one to avoid overlap between the blank areas. When N is 0, there is no need to add abnormality identification in the first picture.

[0081] Optionally, the step S22 comprises:

[0082] Step S221: obtaining a first center point;

[0083] Step S222: determining a first area corresponding to the first center point based on the first center point and the size information, and obtaining color information of the first area, the size of the first area being greater than or equal to the size indicated by the size information, the first area being formed by expanding the corresponding first center point to the surrounding;

[0084] Step S223: in the case where the color information of the first area indicates that the corresponding area is blank, determining the first area as the blank area;

[0085] Step S224: repeating steps S221 to S223 until a target condition occurs, the target condition including one of the following:

[0086] The number of the determined blank areas is N;

[0087] The number of the determined blank areas is less than N, and more blank areas cannot be determined in the first picture.

[0088] In this embodiment, a random point in the first picture is obtained as a first center point, and a first region corresponding to the first center point is obtained by expanding from the position of the first center point in all directions according to the size information, and the size of the first region is greater than or equal to the size indicated by the size information. Then, chrominance information of the first region is obtained, the chrominance information refers to the RGB value of the color in the picture, and whether the first region is blank can be determined according to the chrominance information. In the case where the chrominance information indicates that the first region is blank, the first region is determined as a blank region, which means that adding an abnormal mark to the first region will not overlap with the original line and color content, and will not affect the display of the original picture content; in the case where the chrominance information indicates that the first region is not completely blank, the region cannot be added with an abnormal mark as a blank region, and if an abnormal mark is added, the display of the original picture content will change.

[0089] The way of determining whether the first region is blank according to the chrominance information can be:

[0090] The RGB values of all pixel points in the first region are obtained (sampling can be performed in the case where the first region is too large), the RGB average of the first region is calculated according to the RGB values of all the pixel points, and in the case where the RGB average reaches a preset threshold (for example, the RGB value is greater than or equal to 253), the first region is determined as a blank region.

[0091] Through the above steps S221-S223, a blank region can be determined from the first picture. Steps S221-S223 are repeatedly executed to continuously determine blank regions, and it should be noted that different blank regions do not overlap. In the case where N blank regions are determined, or N blank regions are not determined, but no new blank region that does not overlap with the previous blank region can be determined in the first picture, steps S221-S223 are terminated.

[0092] In this embodiment, N blank regions are sequentially determined by the above method, the execution method is simple, and the overlap between blank regions can be avoided to prevent the overlap of subsequently added abnormal marks, thereby avoiding the failure to identify the abnormal mark, that is, the training effect of the model is improved.

[0093] Optionally, the step S22 includes:

[0094] Step S225: obtaining M different second center points in the first picture, M being an integer greater than or equal to N;

[0095] Step S226: obtaining chrominance information of a first region corresponding to each of the second center points, the size of the first region being greater than or equal to the size indicated by the size information, and the first region being formed by expanding from the corresponding second center point in all directions.

[0096] Step S227: determining N first regions as the blank regions from the M first regions, chrominance information of the N first regions indicating that the corresponding regions are blank.

[0097] In this embodiment, M different second center points are obtained in the first picture in advance, the M second center points are enlarged according to the size information to obtain M first regions, and then N first regions are determined from the M first regions. The N first regions determined from the M first regions need to meet two requirements, one is that the chrominance information corresponding to the first region indicates that the first region is completely blank, and the other is that the first region does not overlap with other first regions. Through the above manner, the M first regions are first confirmed in the first picture, at this time, the positional relationship between the M first regions and whether there is overlap can be directly judged, which facilitates the determination of the N first regions, thereby improving the execution efficiency of the method.

[0098] Optionally, the step S3 comprises:

[0099] Step S31: obtaining the center point of the abnormal identifier;

[0100] Step S32: according to the center point of the abnormal identifier and the N center points corresponding to the N blank regions, the N abnormal identifiers are one-to-one added to the N blank regions to generate the second picture, and the abnormal identifier in the second picture is contained in the corresponding blank region.

[0101] In this embodiment, the center points of the N abnormal identifiers are one-to-one corresponding to the N center points of the N blank regions, so that the N abnormal identifiers are added to the N blank regions to obtain the second picture, which is simple and direct in operation and can improve the execution efficiency of the method to a certain extent.

[0102] Optionally, after the step S4, the method further comprises:

[0103] Step S5: cyclically executing the steps S1-S4 to obtain a training data set containing multiple abnormal identifiers, and the training data set containing multiple abnormal identifiers is used for training an image recognition model.

[0104] In this embodiment, different abnormal identifiers are obtained, and a second picture containing multiple abnormal identifiers is generated to obtain a training data set containing multiple abnormal identifiers. It can be understood that through the above manner, different types of abnormal data can be automatically expanded in different detection scenarios to meet the requirements of the image recognition model training in different scenarios, and the application range of the method of the present embodiment can be expanded.

[0105] Optionally, after the step S5, the method further comprises:

[0106] Step S6: Input the target image into the target model to obtain the image recognition result. The target model is the model obtained by training the image recognition model using the second training dataset.

[0107] In this embodiment, after training the image recognition model using the second training dataset, a target model is obtained. Since the second training dataset has a large amount of data, the target model has a strong ability to identify anomalies. After inputting the target image into the target model, a relatively accurate image recognition result can be obtained.

[0108] like Figure 3 As shown, this embodiment of the invention also provides an apparatus 500 for expanding a training dataset, comprising:

[0109] The acquisition module 501 is used to acquire a first training dataset, an anomaly identifier input by the user, and the size information corresponding to the anomaly identifier. The first training dataset includes a first image.

[0110] The recognition module 502 is used to recognize the first image according to the size information to obtain a blank area of ​​the first image, wherein the size of the blank area is greater than or equal to the size indicated by the size information;

[0111] The generation module 503 is used to generate a second image based on the blank area and the anomaly identifier;

[0112] Add module 504 is used to add the second image to the first training dataset to obtain the second training dataset, which is used to train the image recognition model.

[0113] Optionally, the identification module 502 includes:

[0114] The first submodule is used to obtain a random number N, where N is a natural number.

[0115] The first determining submodule is used to determine N blank areas from the first image based on the size information.

[0116] Optionally, the first determining submodule includes:

[0117] The first acquisition unit is used to acquire the first center point;

[0118] The first determining unit is configured to determine a first region corresponding to the first center point based on the first center point and the size information, and to obtain the chromaticity information of the first region. The size of the first region is greater than or equal to the size indicated by the size information. The first region is formed by expanding outward from the corresponding first center point.

[0119] The second determining unit is configured to determine the first region as the blank region when the chrominance information of the first region indicates that the corresponding region is blank.

[0120] The repeating execution unit is configured to repeatedly execute steps S221 to S223 until a target condition is met, the target condition including one of the following:

[0121] The determined number of the blank regions is N.

[0122] The determined number of the blank regions is less than N, and no more blank regions can be determined in the first picture.

[0123] Optionally, the first determining sub-module includes:

[0124] The second obtaining unit is configured to obtain M different second center points in the first picture, M being an integer greater than or equal to N.

[0125] The third obtaining unit is configured to obtain the chrominance information of a first region corresponding to each second center point, the first region having a size greater than or equal to the size indicated by the size information, the first region being formed by expanding the corresponding second center point in all directions.

[0126] The third determining unit is configured to determine N first regions from the M first regions as the blank regions, the chrominance information of the N first regions indicating that the corresponding regions are blank.

[0127] Optionally, the generating module 503 includes:

[0128] The second obtaining sub-module is configured to obtain the center point of the abnormal identifier.

[0129] The adding sub-module is configured to add N abnormal identifiers to the N blank regions one by one according to the center point of the abnormal identifier and the N center points corresponding to the N blank regions, to generate the second picture, the abnormal identifier being contained in the corresponding blank region in the second picture.

[0130] Optionally, the apparatus 500 is further configured to:

[0131] input a target picture into a target model to obtain a picture recognition result, the target model being a model obtained by training an image recognition model by using a second training data set.

[0132] It should be noted that the apparatus 500 provided by the embodiments of the present application can achieve the advantages of the method 100 as described above. Figure 1The method of expanding the training data set by the computing power of the intelligent computing center shown in the embodiment implements the whole technical process of the method and achieves the same technical effect. To avoid repetition, details are not described here.

[0133] Optionally, as shown in Figure 4 The embodiment of the application also provides an electronic device 600, including a processor 601 and a memory 602, and the memory 602 stores a program or instructions executable on the processor 601. When the program or instructions are executed by the processor 601, the steps of the method for expanding the training data set by the computing power of the intelligent computing center are implemented, and the same technical effect can be achieved. To avoid repetition, details are not described here.

[0134] The application also provides a computer readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the method for developing an agent tool based on the computing power of an intelligent computing center are implemented, and the same technical effect can be achieved. To avoid repetition, details are not described here. The computer readable storage medium includes a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0135] The embodiment of the application also provides a computer program product, including computer instructions. When the computer instructions are executed by a processor, the steps of the method for expanding the training data set by the computing power of the intelligent computing center are implemented, and the same technical effect can be achieved. To avoid repetition, details are not described here. Figure 1

[0136] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or device. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0137] ​Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, also can be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application essentially or say the part of contribution to the prior art can be embodied in the form of software product, the computer software product is stored in a storage medium (such as ROM / RAM, disk, optical disk), including a number of instructions to make a terminal (may be a mobile phone, computer, server, air conditioner, or network equipment, etc.) executes the method described in various embodiments of the present application.

[0138] The embodiments of the present application are described above in conjunction with the drawings, but the present application is not limited to the above-mentioned specific embodiments, the above-mentioned specific embodiments are only illustrative, but not restrictive, those skilled in the art can make many forms under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims, all of which belong to the protection of the present application.

Claims

1. A method for training a data set by computing power expansion of an intelligent computing center, characterized in that, The method comprises: Step S1: obtaining a first training data set, user input abnormality identification, and size information corresponding to the abnormality identification, wherein the first training data set comprises a first picture; Step S2: identifying the first picture according to the size information to obtain a blank area of the first picture, wherein the size of the blank area is greater than or equal to the size indicated by the size information; Step S3: generating a second picture based on the blank area and the abnormality identification; Step S4: replacing the first picture with the second picture in the first training data set to obtain a second training data set, wherein the second training data set is used for training an image recognition model.

2. The method of claim 1, wherein, The step S2 comprises: Step S21: obtaining a random number N, wherein N is a natural number; Step S22: determining N blank areas from the first picture based on the size information.

3. The method of claim 2, wherein, The step S22 comprises: Step S221: obtaining a first center point; Step S222: determining a first area corresponding to the first center point based on the first center point and the size information, and obtaining color information of the first area, wherein the size of the first area is greater than or equal to the size indicated by the size information, and the first area is formed by expanding the corresponding first center point to the surrounding; Step S223: in the case where the color information of the first area indicates that the corresponding area is blank, determining the first area as the blank area; Step S224: repeating steps S221 to S223 until a target condition occurs, wherein the target condition comprises one of the following: The number of determined blank areas is N; The number of determined blank areas is less than N, and no more blank areas can be determined in the first picture.

4. The method of claim 2, wherein, The step S22 comprises: Step S225: obtaining M different second center points in the first picture, wherein M is an integer greater than or equal to N; Step S226: obtaining color information of a first area corresponding to each second center point, wherein the size of the first area is greater than or equal to the size indicated by the size information, and the first area is formed by expanding the corresponding second center point to the surrounding; Step S227: determining N first areas from the M first areas as the blank areas, wherein the color information of the N first areas indicates that the corresponding areas are blank.

5. The method according to claim 2 or 3, characterized in that, The step S3 comprises: Step S31: obtaining a center point of the abnormality identification; Step S32: adding N abnormality identifications to the N blank areas one by one according to the center point of the abnormality identification and N center points corresponding to the N blank areas to generate the second picture, wherein the abnormality identification is contained in the corresponding blank area in the second picture.

6. The method according to any one of claims 1 to 4, characterized in that, After the step S4, the method further comprises: Step S5: cyclically executing the steps S1-S4 to obtain a training data set containing multiple abnormality identifications, wherein the training data set containing multiple abnormality identifications is used for training an image recognition model.

7. An apparatus for expanding a training dataset, the apparatus comprising: The device comprises: An acquisition module is configured to acquire a first training data set, user input abnormality identification, and size information corresponding to the abnormality identification, the first training data set including a first picture; An identification module is configured to identify the first picture according to the size information to obtain a blank area of the first picture, the size of the blank area being greater than or equal to the size indicated by the size information; A generation module is configured to generate a second picture based on the blank area and the abnormality identification; An addition module is configured to replace the first picture with the second picture in the first training data set to obtain a second training data set, the second training data set being used to train an image recognition model.

8. An electronic device, comprising: comprising: a processor, a memory, and a program stored on the memory and executable on the processor, the program, when executed by the processor, implementing the steps of the method for expanding a training data set by computing power of an intelligent computing center according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium has stored thereon a computer program, the computer program, when executed by the processor, implementing the steps of the method for expanding a training data set by computing power of an intelligent computing center according to any one of claims 1 to 6.

10. A computer program product, characterised in that, comprising computer instructions, the computer instructions, when executed by the processor, implementing the steps of the method for expanding a training data set by computing power of an intelligent computing center according to any one of claims 1 to 6.

Images