Image data screening method and related device based on KL divergence and target detection

Abstract

This application discloses a method and related device for image data screening based on KL divergence and target detection. The method includes: obtaining image data containing reference data and data to be screened and performing preprocessing; calculating the KL divergence between each sample in the data to be screened and the entire reference data, and setting a threshold to screen out candidate data; using the reference data to train a target detection model, performing target detection on the candidate data and obtaining a confidence score; setting a confidence score threshold to screen out qualified data. By combining KL divergence with target detection, the present invention effectively screens out image data that is similar to the reference data and has a high target detection confidence score, thereby improving the accuracy and efficiency of image data screening. The method is suitable for large-scale image data screening scenarios and provides higher-quality datasets for tasks such as medical image processing.

Description

Technical Field

[0001] The present invention relates to the field of image processing and artificial intelligence technology, and in particular to an image data screening method based on KL divergence and target detection and a related device. Background Art

[0002] In the field of image data processing, image data screening is an important part of data preprocessing. Especially in large-scale image datasets, screening out high-quality images that meet specific standards is crucial for subsequent analysis and application. At present, common image data screening methods mainly rely on simple pixel thresholds or manual annotations. These methods are inefficient and difficult to adapt to complex and changing image features. In addition, with the rapid increase in the amount of image data, traditional screening methods face huge challenges in accuracy and timeliness, and it is difficult to meet the needs of high-precision image analysis. For example, in target detection tasks, a large number of noisy images or images that differ greatly from the benchmark data will interfere with model training and reduce detection accuracy. Therefore, there is an urgent need for an efficient and automated image data screening technology that can accurately identify images that are similar to the benchmark data and have good target detection effects, so as to improve the quality and efficiency of image data screening and provide reliable data support for subsequent image analysis and applications. Summary of the Invention

[0003] To solve the above problems, the present invention relates to an image data screening method based on KL divergence and target detection and related devices, which include but are not limited to an image data screening device based on KL divergence and target detection, an electronic device, a computer-readable storage medium and a computer program product.

[0004] First, a method for image data screening based on KL divergence and target detection is proposed, which includes the following steps:

[0005] Acquiring and preprocessing image data, wherein the image data includes reference data and data to be screened, wherein the reference data is a reference standard for screening the image data;

[0006] Calculating the KL divergence between each sample in the to-be-screened data and the entire benchmark data;

[0007] Setting a KL divergence threshold, and retaining the to-be-screened data whose KL divergence is lower than the KL divergence threshold as candidate data;

[0008] Training a first target detection model with the benchmark data to obtain a second target detection model, and performing target detection on the candidate data with the second target detection model to obtain a confidence score of the candidate data;

[0009] A confidence threshold is set, and the candidate data with a confidence higher than the confidence threshold is retained to obtain qualified data.

[0010] In conjunction with any embodiment of the present application, calculating the KL divergence between each sample in the to-be-screened data and the entire benchmark data includes:

[0011] Obtaining pixel value distribution of the entire benchmark data;

[0012] Calculating the pixel value distribution of each sample in the data to be screened;

[0013] Calculating the KL divergence between each sample in the data to be screened and the entire benchmark data based on the pixel value distribution of each sample in the data to be screened and the pixel value distribution of the entire benchmark data;

[0014] The pixel value distribution removes irrelevant pixel values, and the irrelevant pixel values ​​are pixel values ​​in the interval [0, 10]∪[245, 255].

[0015] In combination with any embodiment of the present application, the KL divergence threshold is less than or equal to 1.2 times the sample average KL divergence of the data to be screened; or the data to be screened is divided into two groups of high and low divergence through a clustering algorithm, and the boundary value between the two groups is used as the threshold.

[0016] In conjunction with any embodiment of the present application, the target detection includes:

[0017] Performing multi-scale feature extraction on the candidate data to generate a feature map containing spatial location information and semantic information;

[0018] A candidate region is generated based on the feature map, and bounding box regression and category prediction are performed on the candidate region.

[0019] In conjunction with any embodiment of the present application, setting the confidence threshold includes:

[0020] Setting an initial confidence threshold, training a first object detection model with the benchmark data, and obtaining a model accuracy; if the model accuracy is greater than 90%, setting the initial confidence threshold to 90%; if the model accuracy is less than 90%, setting the initial confidence threshold to 85%;

[0021] Normalizing the KL divergence to obtain a normalized KL divergence;

[0022] The confidence threshold formula is: ,

[0023] in, is the confidence threshold, is the initial confidence threshold, is the normalized KL divergence.

[0024] In combination with any embodiment of the present application, the arrangement order of the samples in the qualified data is the same as the arrangement order of the data to be screened.

[0025] In a second aspect, a device for screening image data based on KL divergence and target detection is provided, the device comprising:

[0026] Data acquisition unit: used for acquiring image data and performing preprocessing, wherein the image data includes reference data and data to be screened;

[0027] A KL divergence calculation unit is configured to calculate the KL divergence between each sample in the to-be-screened data and the entire reference data; and is further configured to set a KL divergence threshold value, and retain the to-be-screened data having a KL divergence lower than the KL divergence threshold value as candidate data;

[0028] Target detection unit: used to train a first target detection model with the reference data to obtain a second target detection model, and perform target detection on the candidate data with the second target detection model to obtain the confidence level of the candidate data;

[0029] Confidence screening unit: used to set a confidence threshold, retain the candidate data with confidence higher than the confidence threshold, and obtain qualified data.

[0030] In a third aspect, an electronic device is provided, comprising: a processor, a communication module, a sensor, a user interface, and a storage unit, wherein the storage unit is configured to store computer program code, wherein the program code comprises computer instructions. When the processor executes these instructions, the electronic device performs the method described in the second aspect and any embodiment thereof.

[0031] In a fourth aspect, another electronic device is provided, comprising: a processor, a wireless communication module, a touch screen, a speaker, and a storage unit, wherein the storage unit is configured to store computer program code, wherein the program code comprises computer instructions. When the processor executes these instructions, the electronic device performs the method described in the second aspect and any embodiment thereof.

[0032] In a fifth aspect, a computer-readable storage medium is provided, wherein a computer program is stored, wherein the program includes program instructions. When these instructions are executed by a processor, the processor will perform the method described in the second aspect and any embodiment thereof.

[0033] In a sixth aspect, a computer program product is provided, wherein the computer program product comprises a computer program or instructions. When the computer program or instructions are run on a computer, the computer will execute the method described in the second aspect and any embodiment thereof.

[0034] It should be understood that the above general description and the following detailed description are only used as examples and explanations and do not limit the present application in any way.

[0035] This patent proposes a method for screening image data based on KL divergence and target detection. Its operating process includes: first, acquiring image data and preprocessing it, calculating the KL divergence of each sample in the data to be screened with the benchmark data, and screening out candidate data by setting a threshold; then using the benchmark data to train the target detection model, performing target detection on the candidate data, and further screening out qualified data based on the confidence threshold. Compared with the existing technology, the uniqueness of this patent lies in combining KL divergence and target detection, which not only quantifies the similarity between the image and the benchmark data at the pixel distribution level, but also evaluates the effectiveness of the image through the target detection model. This dual screening mechanism can effectively remove noisy data and images that differ greatly from the benchmark, while retaining high-quality target images, significantly improving the accuracy and efficiency of image data screening, and solving the problem of low efficiency and insufficient precision of traditional methods relying on manual labeling or simple pixel threshold screening, providing an innovative solution for the automated processing of image data. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the background technology, the drawings required for use in the embodiments of the present application or the background technology will be described below.

[0037] The drawings herein are incorporated into and constitute a part of the specification. These drawings illustrate embodiments consistent with the present application and, together with the specification, are used to illustrate the technical solutions of the present application.

[0038] Figure 1 A flowchart of an image data screening method based on KL divergence and target detection provided in an embodiment of the present application;

[0039] Figure 2 This is an architecture diagram of an image data screening device based on KL divergence and target detection provided in an embodiment of the present application;

[0040] Figure 3 A schematic diagram of the hardware architecture of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to allow professionals in this technical field to more fully understand the technical solution of the present application, the technical solution of the present application will be explained in detail and clearly with the help of the accompanying drawings. It should be noted that the described embodiments are only some examples of the present application and do not represent all. Based on these embodiments, those skilled in the art can directly deduce all other possible implementation plans without engaging in creative thinking, and these are also included in the scope of protection of the present application.

[0042] In the specification, claims, and related drawings of this application, the terms "first," "second," and the like are used solely to distinguish between different elements and do not imply any particular order. Furthermore, the use of "including," "having," and their variations denotes non-exclusive inclusion. This means that if a process, method, system, product, or apparatus includes a series of steps or components, the process, method, system, product, or apparatus is not limited to the enumerated steps or components and may include other steps or components not listed, or other steps or units inherent to the process, method, system, product, or apparatus.

[0043] The “embodiment” mentioned in this document refers to any instance in which a particular feature, structure or characteristic is combined, and these instances may belong to at least one embodiment of the present application. The “embodiment” mentioned in this document does not necessarily refer to the same specific case, nor does it mean that they are independent or exclusive alternatives. It should be understood by those skilled in the art that the embodiments described herein can be used in conjunction with other embodiments. It should be understood that in this application, “at least one” includes one or more instances, “a plurality” means two or more instances, and “at least two” means two or more instances.

[0044] It should be understood that the method embodiment of the present application can also be implemented by a processor executing computer program code. The embodiment of the present application is described below in conjunction with the drawings in the embodiment of the present application.

[0045] See also Figure 1 , Figure 1 A flowchart of an image data screening method based on KL divergence and target detection provided in an embodiment of the present application.

[0046] 101. Data acquisition and preprocessing.

[0047] Image data is acquired and preprocessed, wherein the image data includes reference data and data to be screened, and the reference data is a reference standard for screening the image data.

[0048] In this embodiment, the preprocessing of the medical image data at least includes data scaling, cropping, etc.

[0049] 102. Calculate KL divergence.

[0050] The KL divergence between each sample in the to-be-screened data and the entire benchmark data is calculated.

[0051] In this embodiment, the calculation of the KL divergence is based on the pixel value distribution between the sample and the entire reference data.

[0052] In this embodiment, irrelevant pixel values ​​are removed from the pixel value distribution, and the irrelevant pixel values ​​are pixel values ​​in the interval [0, 10]∪[245, 255].

[0053] In another possible implementation, the irrelevant pixel value may vary depending on the task and may be set automatically.

[0054] In another possible implementation, the KL divergence may be calculated by numerical integration to improve the accuracy of the calculation. Numerical integration may employ methods such as the Simpson formula and the trapezoidal formula to approximate the integral expression of the KL divergence.

[0055] In another possible implementation, the calculation of the KL divergence can be combined with the semantic information of the image. The image can be divided into different semantic regions through a pre-trained semantic segmentation model, and then the KL divergence of each semantic region is calculated separately to more accurately measure the difference between images.

[0056] In another possible implementation, the calculation may use other similarity information, such as Euclidean distance, cosine similarity, etc.

[0057] 103. Initial screening.

[0058] A KL divergence threshold is set, and the data to be screened whose KL divergence is lower than the KL divergence threshold is retained as candidate data.

[0059] In this embodiment, the KL divergence threshold is less than or equal to 1.2 times the average KL divergence of the data samples to be screened; or the data to be screened is divided into two groups of high and low divergence through a clustering algorithm, and the boundary value between the two groups is used as the threshold.

[0060] In another possible implementation, the KL divergence threshold value may be dynamically adjusted based on the distribution of the image data. For example, the image data may be divided into different categories using a clustering algorithm, and then different threshold values ​​may be set based on the KL divergence distribution of each category.

[0061] In another possible implementation, the KL divergence threshold may be set based on image category information, with different thresholds being set for images of different categories. For example, a lower threshold may be set for important images of certain categories to retain more candidate data.

[0062] In another possible implementation, the process of screening candidate data can be combined with image similarity information, not only considering KL divergence, but also calculating the similarity between images, such as Euclidean distance, cosine similarity, etc., and comprehensively considering KL divergence and similarity to screen candidate data.

[0063] 104. Target detection model training and testing.

[0064] The first target detection model is trained with the benchmark data to obtain a second target detection model, and the second target detection model is used to perform target detection on the candidate data to obtain the confidence of the candidate data.

[0065] In the embodiment of the present application, the target detection model adopts the YOLO series algorithm, and the version of the YOLO algorithm is no earlier than YOLOv5.

[0066] In another possible implementation, other single-stage target detection algorithms may be used to perform the target region detection.

[0067] In another possible implementation, the training of the first target detection model may adopt a transfer learning method to transfer the pre-trained model to the current task; the pre-trained model may be a general model trained on benchmark data or data similar to the benchmark data.

[0068] 105. Second screening, output data.

[0069] A confidence threshold is set, and the candidate data with a confidence higher than the confidence threshold is retained to obtain qualified data.

[0070] In this embodiment, the setting of the confidence threshold includes: setting an initial confidence threshold, training the first target detection model with the benchmark data, and obtaining the model accuracy; if the model accuracy is greater than 90%, the initial confidence threshold is set to 90%; if the model accuracy is less than 90%, the initial confidence threshold is set to 85%;

[0071] Normalizing the KL divergence to obtain a normalized KL divergence;

[0072] The confidence threshold formula is:

[0073] ,

[0074] in, is the confidence threshold, is the initial confidence threshold, is the normalized KL divergence.

[0075] In another possible implementation method, the confidence threshold may be directly set to 90%.

[0076] In another possible implementation, the process of screening qualified data may be combined with the contextual information of the target, taking into account not only the confidence of the target but also the contextual information around the target, such as the size, position, shape, etc. of the target.

[0077] In another possible implementation method, a manual review process is introduced to review the screening results and ensure data quality.

[0078] Those skilled in the art will understand that in the above-mentioned method of the specific implementation method, the writing order of each step does not mean a strict execution order and does not constitute any limitation on the implementation process. The specific execution order of each step should be determined by its function and possible internal logic.

[0079] The above describes in detail the method of the embodiment of the present application, and the following provides an apparatus of the embodiment of the present application.

[0080] See also Figure 2 , Figure 2 This is a schematic diagram of the structure of an image data screening device based on KL divergence and target detection provided in an embodiment of the present application. The screening device 1 includes a data acquisition unit 11, a KL divergence calculation unit 12, a target detection unit 13, and a confidence screening unit 14. Specifically:

[0081] The data acquisition unit 11 is used to acquire image data and perform preprocessing, wherein the image data includes reference data and data to be screened;

[0082] The KL divergence calculation unit 12 is used to calculate the KL divergence between each sample in the to-be-screened data and the entire reference data; and is also used to set a KL divergence threshold, and retain the to-be-screened data with a KL divergence lower than the KL divergence threshold as candidate data;

[0083] The target detection unit 13 is configured to train a first target detection model using the reference data to obtain a second target detection model, and perform target detection on the candidate data using the second target detection model to obtain the confidence level of the candidate data;

[0084] The confidence screening unit 14 is used to set a confidence threshold, retain the candidate data with a confidence higher than the confidence threshold, and obtain qualified data.

[0085] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0086] See also Figure 3 , Figure 3 The following is a schematic diagram of the hardware architecture of an electronic device described in an embodiment of the present application. The electronic device 2 is primarily composed of a processor 21 and a memory 22. In addition, the device may also include an input device 23 and an output device 24. The processor 21, memory 22, input device 23, and output device 24 are interconnected via connecting components, which may be various interfaces, data cables, or communication buses, and are not specifically specified in the present embodiment.

[0087] Processor 21 may be one or more graphics processing units (GPUs). If processor 21 is a GPU, the GPU may be single-core or multi-core. Optionally, processor 21 may comprise a processor group consisting of multiple GPUs, interconnected via one or more buses. Furthermore, the processor may be other types of processors, which are not specifically limited in this embodiment of the present application.

[0088] Memory 22 is designed to store computer program instructions and various program codes required to execute the present invention. Optionally, the memory may include, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or compact disc read-only memory (CD-ROM), which are used to store relevant instructions and data.

[0089] The input device 23 is used to input data and / or signals, and the output device 24 is used to output data and / or signals. The input device 23 and the output device 24 can be independent devices or an integrated device.

[0090] It should be appreciated that in the embodiment of the present application, the memory 22 can store not only relevant instructions but also relevant data. The embodiment of the present application does not specify the specific data content stored in the memory.

[0091] You should understand that Figure 3 Only a simplified design of an electronic device is shown. In actual use, the electronic device may also include other necessary components, such as different numbers of input / output devices, processors, memories, etc. All electronic devices that can implement the embodiments of this application are within the scope of protection of this application.

[0092] Those skilled in the art will recognize that, according to the components and algorithm steps of each example described in the embodiments disclosed herein, these functions can be implemented by electronic hardware or by combining computer software and electronic hardware. Whether these functions are performed by hardware or software will be determined based on the specific application requirements and design limitations of the technical solution. Technicians can adopt different implementation methods according to the requirements of each specific application, but such implementation methods should not be considered to exceed the scope of protection of this application.

[0093] Professionals should understand that, for the sake of ease of description and simplification, the specific operating procedures of the above-mentioned systems, devices, and components can refer to the corresponding steps in the previous method embodiments and will not be repeated here. At the same time, professionals should also understand that each embodiment in this application has its own focus. For the sake of ease of description and simplification, the same or similar content may not be repeated in different embodiments. Therefore, if a part is not mentioned or not explained in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0094] In the several embodiments provided in this application, it should be recognized that the disclosed systems, devices and methods can also be implemented in other ways. For example, the device embodiments described are only exemplary, in which the division of the units is only a division of logical functions, and there may be different division methods in actual implementation. For example, multiple units or components may be merged or integrated into another system, or certain features may be omitted, or certain steps may not be performed. In addition, the connections between each other shown or discussed, whether direct or indirect, whether coupling or communication connection, may be implemented in electrical, mechanical or other forms through interfaces, devices or units.

[0095] Units described as independent components may or may not actually be physically separate; parts presented as units may or may not be physical entities; that is, they may be centralized in one location or distributed across multiple network nodes. Depending on actual needs, some or all of these units may be selected to achieve the objectives of this embodiment.

[0096] Furthermore, in the various embodiments of the present application, the various functional units may be integrated into a single processing unit, physically exist independently, or two or more units may be combined into a single unit. In the aforementioned embodiments, the relevant functions may be implemented in whole or in part through software, hardware, firmware, or any combination thereof. If software implementation is chosen, it may be implemented in whole or in part in the form of a computer program product. This computer program product comprises one or more computer instructions. When these instructions are loaded and executed on a computer, they will generate, in whole or in part, the processes or functions described in the embodiments of this application. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. These computer instructions may be stored in a computer-readable storage medium or transmitted via such a medium. The computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic cable, DSL) or wireless (e.g., infrared, wireless, microwave, etc.) means. A computer-readable storage medium may be any computer-accessible, usable medium, or a data storage facility such as a server or data center that integrates one or more usable media. These available media may include magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., DVDs), semiconductor media (e.g., SSDs), etc. Those skilled in the art will appreciate that all or part of the process steps for implementing the above-described method embodiments can be accomplished through hardware associated with computer program instructions. These programs can be stored on computer-readable storage media. When executed, these programs will contain the processes for each of the above-described method embodiments. These storage media include, but are not limited to, various media capable of storing program code, such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

Claims

1. A method for screening image data based on KL divergence and target detection, characterized in that: The steps include: Acquiring and preprocessing image data, wherein the image data includes reference data and data to be screened, wherein the reference data is a reference standard for screening the image data; Calculating the pixel value distribution of the entire benchmark data and the pixel value distribution of each sample in the data to be screened; Calculating the KL divergence between each sample in the data to be screened and the entire reference data based on the pixel value distribution of each sample in the data to be screened and the pixel value distribution of the entire reference data; Setting a KL divergence threshold, and retaining the to-be-screened data whose KL divergence is lower than the KL divergence threshold as candidate data; Training a first target detection model with the benchmark data to obtain a second target detection model, and performing target detection on the candidate data with the second target detection model to obtain a confidence score of the candidate data; A first target detection model is trained using the benchmark data to obtain model accuracy, an initial confidence threshold is set based on the model accuracy, the KL divergence is normalized to obtain a normalized KL divergence, a confidence threshold is set based on the initial confidence threshold and the normalized KL divergence, and the candidate data having a confidence higher than the confidence threshold is retained to obtain qualified data; if the model accuracy is greater than 90%, the initial confidence threshold is set to 90%; if the model accuracy is less than 90%, the initial confidence threshold is set to 85%.

2. The method according to claim 1, wherein The pixel value distribution of each sample in the to-be-screened data removes irrelevant pixel values. Similarly, the pixel value distribution of the entire benchmark data also removes irrelevant pixel values. The irrelevant pixel values ​​are pixel values ​​in the interval [0, 10] ∪ [245, 255].

3. The method according to claim 1, wherein The KL divergence threshold is less than or equal to 1.2 times the sample average KL divergence of the data to be screened; or the data to be screened is divided into two groups of high divergence and low divergence by a clustering algorithm, and the boundary value between the two groups is used as the threshold.

4. The method according to claim 1, wherein The target detection includes: Performing multi-scale feature extraction on the candidate data to generate a feature map containing spatial location information and semantic information; A candidate region is generated based on the feature map, and bounding box regression and category prediction are performed on the candidate region.

5. The method according to claim 1, wherein The confidence threshold formula is: ， in, is the confidence threshold, is the normalized KL divergence, is the initial confidence threshold.

6. The method according to claim 1, wherein The order of arrangement of samples in the qualified data is the same as the order of arrangement of the data to be screened.

7. An image data screening device based on KL divergence and target detection, characterized in that: include: Data acquisition unit: acquires image data and performs preprocessing, wherein the image data includes reference data and data to be screened, wherein the reference data is a reference standard for screening the image data; KL divergence calculation unit: used to calculate the pixel value distribution of the entire benchmark data and the pixel value distribution of each sample in the data to be screened; Calculating the KL divergence between each sample in the data to be screened and the overall reference data based on the pixel value distribution of each sample in the data to be screened and the pixel value distribution of the overall reference data; and setting a KL divergence threshold, retaining the data to be screened with a KL divergence lower than the KL divergence threshold as candidate data; Target detection unit: used to train a first target detection model with the reference data to obtain a second target detection model, and perform target detection on the candidate data with the second target detection model to obtain the confidence level of the candidate data; A confidence screening unit is configured to train a first target detection model using the benchmark data to obtain model accuracy, set an initial confidence threshold based on the model accuracy, normalize the KL divergence to obtain a normalized KL divergence, set a confidence threshold based on the initial confidence threshold and the normalized KL divergence, retain the candidate data having a confidence higher than the confidence threshold, and obtain qualified data; if the model accuracy is greater than 90%, the initial confidence threshold is set to 90%; if the model accuracy is less than 90%, the initial confidence threshold is set to 85%.

8. An electronic device, characterized in that: include: A processor and a storage unit, the storage unit is used to store computer program code, the code includes computer instructions, when the processor executes these instructions, the electronic device performs the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the processor is caused to execute the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is run on a computer, the computer is caused to perform the method according to any one of claims 1 to 6.

Images