Training method of palm image quality evaluation model and quality evaluation method

Abstract

The invention discloses a palm image quality evaluation model training method and a quality evaluation method, and belongs to the field of image processing. The method comprises the following steps: acquiring a plurality of palm image samples; classifying the plurality of palm image samples through a classifier, and determining a classification result of each palm image sample; determining the classification accuracy of each palm image sample according to the classification result of the plurality of palm image samples; determining a quality label of the corresponding palm image sample according to the classification accuracy of each palm image sample; evaluating the plurality of palm image samples through a palm image quality evaluation model to obtain a quality score of each palm image sample, the quality scores representing the image quality of the palm image samples; and constructing a first loss function according to the quality label and the quality score of each palm image sample, and training a palm image quality evaluation model based on the first loss function. According to the invention, the practicability of quality evaluation of the palm image is improved.

Description

Technical Field

[0001] This application belongs to the field of image processing technology, and in particular relates to a training method and a quality evaluation method for a palm image quality evaluation model. Background Technology

[0002] With technological advancements, palm images have found widespread application in numerous fields such as identity authentication, intelligent security, and medical imaging. In practical applications, the quality of palm images directly impacts the final performance of image recognition systems. Image quality assessment, as a crucial element in ensuring recognition performance, is becoming increasingly important. Current technologies typically focus only on low-level features such as sharpness and contrast to evaluate image quality. This lack of direct relevance to actual tasks makes it difficult to use quality assessment results to guide practical applications of palm images. Summary of the Invention

[0003] This application aims to address at least one of the technical problems existing in related technologies. To this end, this application proposes a training method and a quality evaluation method for a palm image quality evaluation model, thereby improving the practicality of quality evaluation for palm images.

[0004] Firstly, this application provides a method for training a palm image quality assessment model, the method comprising: Acquire multiple palm image samples; The multiple palm image samples are classified using a classifier to determine the classification result of each palm image sample; Based on the classification results of the multiple palm image samples, the classification accuracy of each palm image sample is determined; Based on the classification accuracy of each palm image sample, determine the quality label of the corresponding palm image sample; The multiple palm image samples are evaluated using a palm image quality evaluation model to obtain a quality score for each palm image sample, and the quality score represents the image quality of the palm image sample. Based on the quality labels and quality scores of each palm image sample, a first loss function is constructed, and the palm image quality evaluation model is trained based on the first loss function.

[0005] In the above technical solution, after acquiring palm image samples, a classifier is used to classify them, determining the classification results of each palm image sample. This provides a sufficient data foundation for model training. Based on the classification results of multiple palm image samples, the classification accuracy of each palm image sample is determined, and a corresponding quality label is determined based on the classification accuracy. The quality label converts the numerical classification accuracy into discrete label information, providing supervision information to facilitate the subsequent construction of the loss function and model training. The palm image quality evaluation model evaluates the image samples to obtain the quality score of each palm image sample. A loss function is constructed based on the quality label and quality score to train the model. This allows the quality score to reflect the usefulness of the palm image sample for application tasks such as image classification, realizing a direct correlation between image quality evaluation and image classification. The quality score output by the trained palm image quality evaluation model can be used to better guide the practical application of palm images (such as classifying palm images using a classifier), improving the fit between palm image quality evaluation and actual palm image classification scenarios, and enhancing the practicality of palm image quality evaluation.

[0006] According to some embodiments of this application, determining the classification accuracy of each of the plurality of palm image samples based on the classification results of the palm image samples includes: For each palm image sample, the intra-class score of the palm image sample is calculated based on other palm image samples with the same classification result as the palm image sample. Calculate the inter-class score of the palm image sample based on other palm image samples whose classification results are different from those of the palm image sample; The classification accuracy of the palm image sample is determined based on the intra-class score and the inter-class score.

[0007] According to some embodiments of this application, after training the palm image quality assessment model based on the first loss function, the method further includes: The palm image quality evaluation model is used to evaluate the multiple palm image samples to obtain the quality score of each palm image sample; The weight of each palm image sample is determined based on its quality score. A second loss function is constructed based on the weights of each palm image sample, and the classifier is trained based on the second loss function; The process returns to the step of classifying the multiple palm image samples using a classifier and determining the classification result of each palm image sample, and is executed iteratively until a preset stopping condition is met.

[0008] According to some embodiments of this application, determining the weight of the corresponding palm image sample based on the quality score of each palm image sample includes: Determine the stability factor; For each palm image sample, the weight of the palm image sample is determined based on the quality score of the palm image sample and the stability factor; The weight of the palm image sample is negatively correlated with its quality score.

[0009] According to some embodiments of this application, constructing the second loss function based on the weights of each of the palm image samples includes: For each of the palm image samples, calculate the cross-entropy between the classification result and the classification result label of the palm image sample; The second loss function is constructed by combining the cross-entropy corresponding to each palm image sample with the weight of each palm image sample.

[0010] According to some embodiments of this application, the method further includes: If the number of iterations reaches a preset threshold, the preset stopping condition is determined to be met. If the classification accuracy obtained by verifying the classifier through multiple verification palm image samples converges, it is determined that the preset stopping condition is met; If the quality score obtained by verifying the palm image quality evaluation model through the multiple verified palm image samples converges, it is determined that the preset stopping condition is met.

[0011] According to some embodiments of this application, determining the quality label of the corresponding palm image sample based on the classification accuracy of each palm image sample includes: For each of the palm image samples, if the classification accuracy of the palm image sample is not less than a preset accuracy threshold, the quality label of the image sample is determined to be a positive sample. If the classification accuracy of the palm image sample is less than a preset accuracy threshold, the quality label of the image sample is determined to be a negative sample.

[0012] Secondly, this application provides a method for evaluating the quality of a palm image, the method comprising: Obtain palm image; The palm image is evaluated using a palm image quality assessment model to obtain a quality score for the palm image; the quality score characterizes the image quality of the palm image. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method described in the first aspect above.

[0013] Thirdly, this application provides a training device for a palm image quality evaluation model, the device comprising: The first acquisition module is used to acquire multiple palm image samples; The classification module is used to classify the plurality of palm image samples using a classifier and determine the classification result of each palm image sample; The accuracy determination module is used to determine the classification accuracy of each palm image sample based on the classification results of the plurality of palm image samples; The label determination module is used to determine the quality label of the corresponding palm image sample based on the classification accuracy of each palm image sample. The first evaluation module is used to evaluate the plurality of palm image samples through a palm image quality evaluation model to obtain a quality score for each palm image sample, wherein the quality score characterizes the image quality of the palm image sample. The training module is used to construct a first loss function based on the quality label and quality score of each palm image sample, and to train the palm image quality evaluation model based on the first loss function.

[0014] Fourthly, this application provides a palm image quality evaluation device, the device comprising: The second acquisition module is used to acquire palm images; The second evaluation module is used to evaluate the palm image using a palm image quality evaluation model to obtain a quality score for the palm image; the quality score represents the image quality of the palm image. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method described in the first aspect above.

[0015] Fifthly, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the palm image quality evaluation model training method as described in the first aspect above, or implements the palm image quality evaluation method as described in the second aspect above.

[0016] In a sixth aspect, this application provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the palm image quality evaluation model training method as described in the first aspect above, or implements the palm image quality evaluation method as described in the second aspect above.

[0017] In a seventh aspect, this application provides a chip, the chip including a processor and a communication interface, the communication interface being coupled to the processor, the processor being used to run programs or instructions to implement the palm image quality evaluation model training method as described in the first aspect, or to implement the palm image quality evaluation method as described in the second aspect above.

[0018] Eighthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the palm image quality evaluation model training method as described in the first aspect above, or implements the palm image quality evaluation method as described in the second aspect above.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart illustrating the training method for a palm image quality evaluation model provided in some embodiments of this application; Figure 2 This is a flowchart illustrating a palm image quality evaluation method provided in some embodiments of this application; Figure 3 This is a schematic diagram of the structure of a palm image quality evaluation model training device provided in some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a palm image quality evaluation device provided in some embodiments of this application; Figure 5 These are schematic diagrams of the structure of electronic devices provided in some embodiments of this application. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0022] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0023] The training method and quality evaluation method of the palm image quality evaluation model provided in this application will be described in detail below with reference to the accompanying drawings, through specific embodiments and application scenarios.

[0024] The training method for the palm image quality evaluation model provided in this application embodiment can be executed by an electronic device or a functional module or entity in an electronic device that can implement the training method for the palm image quality evaluation model. The electronic devices mentioned in this application embodiment include, but are not limited to, mobile phones, tablets, computers, cameras, and wearable devices. The training method for the palm image quality evaluation model provided in this application embodiment will be described below using an electronic device as the execution subject.

[0025] Figure 1 This is a flowchart illustrating the training method for a palm image quality assessment model provided in some embodiments of this application. For example... Figure 1 As shown, the training method for the palm image quality evaluation model includes steps 110, 120, 130, 140, 150, and 160.

[0026] Step 110: Obtain multiple palm image samples.

[0027] As is understood, a palm image refers to an image that includes palm information. For example, it could be a palm print image, a palm vein image, or a palm print / palm vein fusion feature image. Palm image samples can come from different acquisition devices, different acquisition environments, and different individuals, and can be used to train palm image quality assessment models or classifiers.

[0028] For example, a palm image sample can be a palmprint and palm vein fusion feature map, which includes palmprint features and palm vein features, representing palmprint information and palm vein information. In applications such as biometrics and identity verification, the palmprint and palm vein fusion feature map has higher accuracy and stability compared to individual palmprint images or palm vein images.

[0029] Step 120: Classify the multiple palm image samples using a classifier to determine the classification result of each palm image sample.

[0030] The classifier can be a machine learning or deep learning model such as a deep convolutional neural network. It takes a palm image sample as input and classifies (or identifies) the palm image sample, determining that it belongs to a predefined category and outputting the classification result. The classification result of the palm image sample can be considered as the category to which the palm image sample belongs.

[0031] For example, in a scenario where palm images are used for identity recognition, the classification result of a palm image sample can be the user identifier of the user corresponding to the palm image sample.

[0032] In some embodiments, the method further includes: obtaining classification result labels corresponding to each of the palm image samples. The classification result labels corresponding to the palm image samples can be used as training labels to perform preliminary training on the classifier, and the preliminarily trained classifier is used to classify multiple palm image samples to determine the classification result of each palm image sample. The following is a specific example of a loss function for the initial training of a classifier:

[0033] in, This is the loss value. Palm image sample The classification result labels, Palm image sample The classification results, where N is the number of palm image samples. To calculate cross-entropy.

[0034] Step 130: Determine the classification accuracy of each palm image sample based on the classification results of the multiple palm image samples.

[0035] Understandably, classification accuracy reflects how accurately the classifier classifies each palm image sample. Furthermore, classification accuracy also reflects the quality of the palm image samples. Specifically, palm image samples with high classification accuracy are generally considered to have good image quality, making them easier for the classifier to classify correctly, thus resulting in a high classification accuracy; conversely, palm image samples with low classification accuracy are considered to have poor image quality, making them difficult for the classifier to classify accurately, thus resulting in a low classification accuracy.

[0036] For example, if pre-labeled classification results exist for the palm image samples, the classification result of the classifier can be compared with the classification result label to determine the classification accuracy. If no classification result label exists, the classification accuracy can be determined based on the similarity of palm image samples with the same classification result or the difference between palm image samples with different classification results. By analyzing the similarity of samples with the same classification result and the difference between samples with different classification results, it is helpful to determine the inherent structure and feature distribution patterns of the palm image data, thereby enabling the determination of classification accuracy based on the characteristics and interrelationships of the palm image samples themselves.

[0037] Step 140: Determine the quality label of the corresponding palm image sample based on the classification accuracy of each palm image sample.

[0038] For example, the quality label of a palm image sample can be determined based on its classification accuracy by setting a classification accuracy threshold or a classification rule. For instance, samples with a classification accuracy higher than a preset threshold can be labeled as high-quality samples, and samples with a classification accuracy lower than a preset threshold can be labeled as low-quality samples. Alternatively, the quality label of a palm image sample can be a binary label, specifically 1 or 0, where 1 represents a positive sample (high-quality) and 0 represents a negative sample (low-quality).

[0039] Understandably, quality labels convert numerical classification accuracy into discrete label information, providing clear identifiers for palm image samples to facilitate subsequent loss function construction and model training. Specifically, quality labels provide supervision information for subsequent model training, enabling the model to learn palm image features under different quality labels.

[0040] Step 150: Evaluate the multiple palm image samples using a palm image quality evaluation model to obtain a quality score for each palm image sample, wherein the quality score characterizes the image quality of the palm image sample.

[0041] The palm image quality assessment model can be a machine learning or deep learning model, such as a deep convolutional neural network. For example, the palm image quality assessment model takes palm image samples as input and learns image features (such as palm texture and palm edges) to output a quality score corresponding to the palm image sample. The quality score can quantitatively represent the quality of the palm image sample, providing a data foundation for subsequently constructing the loss function.

[0042] Step 160: Construct a first loss function based on the quality label and quality score of each palm image sample, and train the palm image quality evaluation model based on the first loss function.

[0043] The first function refers to the loss function of the palm image quality assessment model. This loss function can be a mean squared error loss function or a cross-entropy loss function, etc. For example, the mean squared error or cross-entropy between the quality score and the quality label can be calculated. During training, optimization algorithms (such as stochastic gradient descent, Adam algorithm, etc.) are used to adjust the parameters of the palm image quality assessment model based on the gradient information of the loss function, gradually reducing the value of the loss function and thus making the quality score predicted by the model gradually approach the true quality label.

[0044] The following is a specific example of a first loss function: Let the palm image sample The quality label is a binary label, specifically 1 or 0, where 1 represents a positive sample (high quality) and 0 represents a negative sample (low quality). The first loss function is:

[0045] in, This is the loss value. Palm image sample Quality label, Palm image sample The quality score is N, where N is the number of palm image samples.

[0046] Understandably, the quality label is determined based on the classification results of multiple palm image samples by a classifier. The loss function is calculated based on the quality label and the quality score output by the model, allowing the quality score to reflect the usefulness of the palm image sample for application tasks such as image classification, thus achieving a direct correlation between image quality assessment and image classification. The quality assessment results (quality scores) output by the trained palm image quality assessment model can be used to better guide the practical application of palm images (such as classifying palm images using a classifier), improving the fit between palm image quality assessment and actual application scenarios, helping to reduce biases in practical applications, and enhancing the practicality of palm image quality assessment.

[0047] The palm image quality assessment model training method provided in this application, after acquiring palm image samples, classifies them using a classifier to determine the classification result of each palm image sample, providing a sufficient data foundation for model training. Based on the classification results of multiple palm image samples, the classification accuracy of each palm image sample is determined, and the corresponding quality label is determined based on the classification accuracy. The quality label converts the numerical classification accuracy into discrete label information, providing supervision information to facilitate the subsequent construction of the loss function and model training. The palm image quality assessment model evaluates the image samples to obtain the quality score of each palm image sample, and a loss function is constructed based on the quality label and quality score to train the model. This allows the quality score to reflect the usefulness of the palm image sample for application tasks such as image classification, realizing a direct correlation between image quality assessment and image classification. The quality score output by the trained palm image quality assessment model can be used to better guide the practical application of palm images (such as classifying palm images using a classifier), improving the fit between palm image quality assessment and actual palm image classification scenarios, and enhancing the practicality of palm image quality assessment.

[0048] In some embodiments of this application, determining the classification accuracy of each palm image sample based on the classification results of the plurality of palm image samples includes: For each palm image sample, the intra-class score of the palm image sample is calculated based on other palm image samples with the same classification result as the palm image sample. Calculate the inter-class score of the palm image sample based on other palm image samples whose classification results are different from those of the palm image sample; The classification accuracy of the palm image sample is determined based on the intra-class score and the inter-class score.

[0049] Understandably, the intra-class score measures the similarity or association between a palm image sample and other samples classified into the same category; the inter-class score measures the similarity or association between a palm image sample and other samples classified into different categories.

[0050] For other palm image samples with the same classification result as the palm image sample, calculate the intra-class score of the palm image sample. For example, this can be done by calculating the similarity between the feature vector of the palm image sample and the feature vector of the same type of sample, such as cosine similarity; then averaging all similarities to obtain the intra-class score of the sample, or by calculating the distance between the feature vector of the given sample and the feature vector of the same type of sample, such as Euclidean distance, taking the reciprocal or normalizing it, and then averaging it as the intra-class score.

[0051] Correspondingly, for a given sample, the inter-class score of the sample can be obtained by calculating the difference between the feature vector of the sample and the feature vectors of samples from different classes, and then averaging all similarities.

[0052] The classification accuracy of palm image samples is determined by setting a ratio between intra-class and inter-class scores. In some embodiments, weight parameters can be introduced to adjust the influence of intra-class and inter-class scores on the classification accuracy.

[0053] Here is a specific example: For each classification result Calculate the score within the class. and inter-class scores ; Intraclass score for:

[0054] Inter-class scores for:

[0055] in, For similarity calculation function, , These are palm image samples. Palm image sample The classification results Palm image sample The classification results; For each sample Calculate its classification accuracy:

[0056] in, This is the similarity threshold.

[0057] The palm image quality assessment model training method provided in this application calculates intra-class scores for palm image samples with the same classification results as the palm image samples; calculates inter-class scores for palm image samples with different classification results than the palm image samples; and determines the classification accuracy by combining intra-class and inter-class scores. This method comprehensively considers the similarity between palm image samples and samples of the same class and the differences between samples of different classes, so that the classification accuracy can more accurately reflect the accuracy of the classifier model in classifying a single sample. This provides more valuable feedback information for subsequent training and optimization of the palm image quality assessment model, helps to improve the fit between palm image quality assessment and actual application scenarios of palm images, and thus improves the practicality of palm image quality assessment.

[0058] In some embodiments of this application, after training the palm image quality assessment model based on the first loss function, the method further includes: The palm image quality evaluation model is used to evaluate the multiple palm image samples to obtain the quality score of each palm image sample; The weight of each palm image sample is determined based on its quality score. A second loss function is constructed based on the weights of each palm image sample, and the classifier is trained based on the second loss function; The process returns to the step of classifying the multiple palm image samples using a classifier and determining the classification result of each palm image sample, and is executed iteratively until a preset stopping condition is met.

[0059] It is understandable that the weights of the palm image samples are coefficients assigned to each sample based on the quality score of the palm image sample. The magnitude of the weights reflects the importance of the corresponding palm image sample in the subsequent classifier training process.

[0060] For example, formulas such as linear weighting can be used to determine the weight of a palm image sample based on the ratio of its quality score to the sum of the quality scores of all palm image samples. Alternatively, a preset quality score threshold can be set, and the corresponding weight can be determined based on the threshold at which the quality score of the palm image sample falls.

[0061] A second loss function is constructed based on the weights of each palm image sample. This second loss function is the loss function of the classifier. For example, it can be based on the standard cross-entropy loss function with the weights of each sample introduced, or it can be based on the weighted mean square error loss function with the weights of each sample introduced. By constructing a weighted second loss function, the classifier can focus on the classification error of samples with different quality scores during training, thereby optimizing the parameters of the classifier and improving the classification accuracy of the classifier for samples with different quality scores.

[0062] By iteratively executing the above steps, the palm image quality assessment model and classifier are continuously optimized and improved, enabling the palm image quality assessment model to more accurately evaluate image quality and the classifier to more accurately classify images. This realizes a feedback mechanism between image quality assessment and image classification, and uses image classification feedback to guide image quality assessment, forming an adaptive quality assessment closed-loop system.

[0063] The palm image quality assessment model training method provided in this application evaluates the quality of samples and outputs quality scores. Based on these quality scores, sample weights are determined, enabling the classifier to focus on the classification errors of samples with different quality scores during training. This optimizes the classifier's parameters, resulting in a weighted classifier. The classification results of the weighted classifier are fed back into the palm image quality assessment model, serving as reference information for re-evaluating sample quality. This realizes a feedback mechanism between image quality assessment and image classification, and uses image classification feedback to guide image quality assessment. This further improves the fit between palm image quality assessment and practical application scenarios of palm images, contributing to increased accuracy and practicality in palm image quality assessment.

[0064] In some embodiments of this application, determining the weight of the corresponding palm image sample based on the quality score of each palm image sample includes: Determine the stability factor; For each palm image sample, the weight of the palm image sample is determined based on the quality score of the palm image sample and the stability factor; The weight of the palm image sample is negatively correlated with its quality score.

[0065] It is understood that the stabilization factor is a pre-set parameter used to adjust the negative correlation between the weights and quality scores of palm image samples, ensuring that the weights have a reasonable distribution range under different quality scores. The stabilization factor can be determined based on historical experience or actual needs, and this application does not impose specific limitations on it.

[0066] For example, the weights of palm image samples can be determined using either linear or nonlinear negative correlation calculation methods. The following is a specific example of determining the weights of palm image samples:

[0067] in, Palm image sample The weight, Palm image sample The mass fraction, This is a preset, non-zero stability factor.

[0068] The palm image quality assessment model training method provided in this application embodiment has a negative correlation between the weight of palm image samples and their quality scores. The weight of palm image samples is used to construct a second loss function. Low-quality palm image samples (such as those with quality scores below a preset threshold) have higher weights, which increases their contribution to the loss function. This makes the classifier pay more attention to classifying low-quality samples. It helps the classifier learn the unique features or patterns that may exist in low-quality samples, avoids ignoring low-quality samples due to excessive focus on high-quality samples, and thus improves the classifier's overall classification ability for various types of samples, especially its robustness in classifying low-quality samples.

[0069] In some embodiments of this application, constructing the second loss function based on the weights of each of the palm image samples includes: For each of the palm image samples, calculate the cross-entropy between the classification result and the classification result label of the palm image sample; The second loss function is constructed by combining the cross-entropy corresponding to each palm image sample with the weight of each palm image sample.

[0070] It is understandable that the weights of the palm image samples are coefficients assigned to each sample based on the quality score of the palm image sample. The magnitude of the weights reflects the importance of the corresponding palm image sample in the subsequent classifier training process.

[0071] In practical applications, the quality of input palm images may vary. However, the classifier and quality evaluation model trained with weights can maintain good performance even under poor lighting conditions, blurry images, or occlusion, providing a more reliable guarantee for practical applications.

[0072] Here is a specific example:

[0073] in, This is the loss value. Palm image sample The classification result labels, Palm image sample The classification results, where N is the number of palm image samples. To calculate cross-entropy, Palm image sample The weight.

[0074] The palm image quality assessment model training method provided in this application, by assigning different weights to samples of different qualities when constructing the loss function, can guide the classifier to pay more attention to high-weight samples, such as more challenging low-quality palm image samples, during the training process. This improves the classification accuracy and robustness of the classifier for samples of different qualities, ultimately enabling the classifier to achieve better classification results on all samples. Furthermore, it provides more accurate feedback information for the palm image quality assessment model. Image classification feedback can guide image quality assessment, helping to improve the fit between palm image quality assessment and actual palm image application scenarios.

[0075] In some embodiments of this application, the method further includes: If the number of iterations reaches a preset threshold, the preset stopping condition is determined to be met. If the classification accuracy obtained by verifying the classifier through multiple verification palm image samples converges, it is determined that the preset stopping condition is met; If the quality score obtained by verifying the palm image quality evaluation model through the multiple verified palm image samples converges, it is determined that the preset stopping condition is met.

[0076] The preset iteration threshold is a pre-defined upper limit for the number of iterations. For example, after each iteration, the iteration count is incremented by 1 and compared to the preset threshold. If the iteration count reaches the preset threshold, the preset stopping condition is met. Setting a preset iteration threshold can prevent the palm image quality assessment model or classifier from overfitting to palm image samples, leading to a decrease in generalization ability on new data. By limiting the number of iterations, the complexity of the model can be controlled to some extent, improving its generalization performance.

[0077] In some embodiments, the classification accuracy obtained by validating the classifier using multiple valid palm image samples converges, including: Multiple verification palm image samples are classified using a classifier to determine the classification result of each verification palm image sample; Based on the classification results of the multiple verification palm image samples, the classification accuracy of each verification palm image sample is determined; The verification classification accuracy is determined based on the classification accuracy of each of the verification palm image samples. If the change in the verification classification accuracy corresponding to the current iteration is less than the change threshold of the previous iteration, the classification accuracy is determined to have converged.

[0078] Specifically, if the change in classification accuracy of each validated palm image sample compared to the previous iteration is less than a first threshold during the current iteration, it indicates that the classifier's performance has not significantly improved in the current iteration, and the palm image quality assessment model or classifier may have already converged. Stopping the iteration at this point avoids unnecessary computation and improves the training data. Setting a reasonable threshold can prevent stopping the iteration too early or too late; if the preset threshold is set too low, the model may stop training before it truly converges, affecting the model's final performance; conversely, if the preset threshold is set too high, it may lead to overtraining.

[0079] For example, after each iteration, the classification accuracy of the current iteration can be recorded, and the change in classification accuracy compared to the previous iteration can be calculated and compared with a preset threshold. An example is given below:

[0080] in, It is the first Samples after round of iterations The classification accuracy It is the first Samples after round of iterations The classification accuracy The threshold value for the first change (e.g., 0.001) is N, and N is the number of palm image samples.

[0081] In some embodiments, the convergence of the quality score obtained by validating the palm image quality assessment model using the plurality of valid palm image samples includes: The multiple verified palm image samples are evaluated using a palm image quality evaluation model to obtain a quality score for each of the verified palm image samples. The verification quality score is determined based on the quality score of each of the verification palm image samples. If the change in the verification quality score corresponding to the current iteration is less than the change threshold of the previous iteration, the preset stopping condition is met.

[0082] Specifically, if the change in quality score of each validated palm image sample in the current iteration is less than the second change threshold compared to the previous iteration, it indicates that the palm image quality assessment model has stabilized its evaluation of sample quality. Since the palm image quality assessment model and the classifier are trained collaboratively, the stability of the quality assessment model helps the classifier obtain more accurate quality weight information, thereby improving classification performance. An example is given below:

[0083] in, It is the first Samples in round iteration The mass fraction, It is the first Samples in round iteration The mass fraction, The second threshold for change. This represents the number of samples.

[0084] The palm image quality assessment model training method provided in this application sets three preset stopping conditions and stops iterative execution when at least one of the three preset stopping conditions is met. The preset number threshold limits the maximum number of iterations, which can effectively prevent the model from overfitting on the training data. When the classification accuracy or quality score has stabilized, continuing to iterate has limited improvement on the model performance. This method improves training efficiency while ensuring the training effect of the palm image quality assessment model and classifier, and avoids overfitting of the palm image quality assessment model and classifier and unnecessary consumption of computing resources.

[0085] In some embodiments of this application, determining the quality label of the corresponding palm image sample based on the classification accuracy of each palm image sample includes: For each of the palm image samples, if the classification accuracy of the palm image sample is not less than a preset accuracy threshold, the quality label of the image sample is determined to be a positive sample. If the classification accuracy of the palm image sample is less than a preset accuracy threshold, the quality label of the image sample is determined to be a negative sample.

[0086] It is understandable that defining samples with a classification accuracy of not less than a preset threshold as positive samples can filter out palm image samples with reliable classification results and high quality. Positive samples can provide high-quality training data for the palm image quality assessment model, which helps the palm image quality assessment model learn the features of high-quality palm images. By identifying samples with a classification accuracy below a preset threshold as negative samples, we can filter out palm image samples with unreliable model classification results and low quality. Negative samples can provide negative examples for model training, helping the model distinguish the feature differences between high-quality and low-quality palm images, thereby improving the model's ability to recognize low-quality palm images and enhancing the robustness of the palm image quality evaluation model.

[0087] Here is a specific example: Set a preset accuracy threshold ; exist In the case of determining palm image samples For high-quality positive samples: exist In the case of determining palm image samples These are low-quality negative samples; in, Palm image sample The classification accuracy.

[0088] The palm image quality assessment model training method provided in this application determines the corresponding quality label based on the classification accuracy of each palm image sample, realizing the division of positive and negative samples. It can provide accurate and targeted labeled data for the training of the palm image quality assessment model, guide the model to learn the feature differences of images of different quality, and thus improve the fit between the palm image quality assessment model and the actual classification scenario of palm images.

[0089] Figure 2 This is a flowchart illustrating a palm image quality evaluation method provided in some embodiments of this application. For example... Figure 2 As shown, the palm image quality evaluation method includes steps 210 and 220.

[0090] Step 210: Obtain the palm image.

[0091] As is understood, a palm image refers to an image that includes palm information. For example, it can be a palm print image, a palm vein image, or a palm print / palm vein fusion feature image. Palm image samples can come from different acquisition devices, different acquisition environments, and different individuals, and can be used in various scenarios such as identity authentication and intelligent security.

[0092] Step 220: Evaluate the palm image using a palm image quality assessment model to obtain a quality score for the palm image; the quality score characterizes the image quality of the palm image.

[0093] The palm image quality evaluation model is trained based on the aforementioned palm image quality evaluation model training method.

[0094] The palm image quality assessment model can be a machine learning or deep learning model, such as a deep convolutional neural network. For example, the palm image quality assessment model takes palm image samples as input and learns image features (such as palm texture and palm edges) to output a quality score corresponding to the palm image sample. The quality score can quantitatively represent the quality level of the palm image sample.

[0095] In some embodiments, the palm image quality assessment model is trained through the following steps: Acquire multiple palm image samples; The multiple palm image samples are classified using a classifier to determine the classification result of each palm image sample; Based on the classification results of the multiple palm image samples, the classification accuracy of each palm image sample is determined; Based on the classification accuracy of each palm image sample, determine the quality label of the corresponding palm image sample; The multiple palm image samples are evaluated using a palm image quality evaluation model to obtain a quality score for each palm image sample, and the quality score represents the image quality of the palm image sample. Based on the quality labels and quality scores of each palm image sample, a first loss function is constructed, and the palm image quality evaluation model is trained based on the first loss function.

[0096] The palm image quality assessment method provided in this application acquires a palm image and evaluates the palm image using a palm image quality assessment model to obtain a quality score characterizing the image quality. The palm image quality assessment model is a pre-trained model, and the output quality score can reflect the usefulness of the palm image sample for application tasks such as image classification, realizing a direct correlation between image quality assessment and image classification. The quality score output by the trained palm image quality assessment model can be used to better guide the practical application of palm images (such as classifying palm images using a classifier), improving the fit between palm image quality assessment and actual palm image classification scenarios, and enhancing the practicality of palm image quality assessment.

[0097] The palm image quality assessment model training method provided in this application can be executed by a palm image quality assessment model training device. This application uses the palm image quality assessment model training device executing the palm image quality assessment model training method as an example to illustrate the palm image quality assessment model training device provided in this application.

[0098] Figure 3 This is a schematic diagram of the structure of a palm image quality evaluation model training device provided in some embodiments of this application. For example... Figure 3 As shown, the palm image quality evaluation model training device 300 includes: The first acquisition module 301 is used to acquire multiple palm image samples; The classification module 302 is used to classify the plurality of palm image samples by a classifier and determine the classification result of each palm image sample; The accuracy determination module 303 is used to determine the classification accuracy of each palm image sample based on the classification results of the plurality of palm image samples; The label determination module 304 is used to determine the quality label of the corresponding palm image sample based on the classification accuracy of each palm image sample. The first evaluation module 305 is used to evaluate the plurality of palm image samples through a palm image quality evaluation model to obtain a quality score for each palm image sample, wherein the quality score characterizes the image quality of the palm image sample. The training module 306 is used to construct a first loss function based on the quality label and quality score of each palm image sample, and to train the palm image quality evaluation model based on the first loss function.

[0099] In some embodiments, the palm image quality evaluation model training device 300 further includes an iteration module, which is used for: The palm image quality evaluation model is used to evaluate the multiple palm image samples to obtain the quality score of each palm image sample; The weight of each palm image sample is determined based on its quality score. A second loss function is constructed based on the weights of each palm image sample, and the classifier is trained based on the second loss function; The process returns to the step of classifying the multiple palm image samples using a classifier and determining the classification result of each palm image sample, and is executed iteratively until a preset stopping condition is met.

[0100] In some embodiments, the palm image quality evaluation model training device 300 further includes a stopping condition determination module, which is used for: If the number of iterations reaches a preset threshold, the preset stopping condition is determined to be met. If the classification accuracy obtained by verifying the classifier through multiple verification palm image samples converges, it is determined that the preset stopping condition is met; If the quality score obtained by verifying the palm image quality evaluation model through the multiple verified palm image samples converges, it is determined that the preset stopping condition is met.

[0101] In some embodiments, determining the weight of a corresponding palm image sample based on the quality score of each palm image sample includes: Determine the stability factor; For each palm image sample, the weight of the palm image sample is determined based on the quality score of the palm image sample and the stability factor; The weight of the palm image sample is negatively correlated with its quality score.

[0102] In some embodiments, constructing the second loss function based on the weights of each of the palm image samples includes: For each of the palm image samples, calculate the cross-entropy between the classification result and the classification result label of the palm image sample; The second loss function is constructed by combining the cross-entropy corresponding to each palm image sample with the weight of each palm image sample.

[0103] In some embodiments, the accuracy determination module 303 is used for: For each palm image sample, the intra-class score of the palm image sample is calculated based on other palm image samples with the same classification result as the palm image sample. Calculate the inter-class score of the palm image sample based on other palm image samples whose classification results are different from those of the palm image sample; The classification accuracy of the palm image sample is determined based on the intra-class score and the inter-class score.

[0104] In some embodiments, the label determination module 304 is used for: For each of the palm image samples, if the classification accuracy of the palm image sample is not less than a preset accuracy threshold, the quality label of the image sample is determined to be a positive sample. If the classification accuracy of the palm image sample is less than a preset accuracy threshold, the quality label of the image sample is determined to be a negative sample.

[0105] The handheld image quality evaluation model training device in this application embodiment can be an electronic device or a component within an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices besides a terminal. For example, the electronic device can be a mobile phone, tablet computer, laptop computer, handheld computer, in-vehicle electronic device, mobile internet device (MID), augmented reality (AR) / virtual reality (VR) device, robot, wearable device, ultra-mobile personal computer (UMPC), netbook, or personal digital assistant (PDA), etc. It can also be a server, network attached storage (NAS), personal computer (PC), television (TV), ATM, or self-service machine, etc. This application embodiment does not specifically limit the specific devices.

[0106] The palm image quality evaluation model training device in this application embodiment can be a device with an operating system. This operating system can be a Microsoft (Windows) operating system, an Android operating system, an iOS operating system, or other possible operating systems; this application embodiment does not specifically limit it.

[0107] The palm image quality evaluation model training device provided in this application embodiment can realize all the processes implemented in the above palm image quality evaluation model training method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0108] The palm image quality evaluation method provided in this application can be executed by a palm image quality evaluation device. This application uses the palm image quality evaluation device executing the palm image quality evaluation method as an example to illustrate the palm image quality evaluation device provided in this application.

[0109] Figure 4 This is a schematic diagram of the structure of a palm image quality evaluation device provided in some embodiments of this application. For example... Figure 4 As shown, the palm image quality evaluation device 400 includes: The second acquisition module 401 is used to acquire a palm image; The second evaluation module 402 is used to evaluate the palm image using a palm image quality evaluation model to obtain a quality score for the palm image; the quality score represents the image quality of the palm image. The palm image quality evaluation model is trained based on the aforementioned palm image quality evaluation model training method.

[0110] The palm image quality evaluation device provided in this application embodiment can realize all the processes implemented in the above palm image quality evaluation method embodiment and achieve the same technical effect. To avoid repetition, it will not be described again here.

[0111] Figure 5 These are schematic diagrams of the structure of electronic devices provided in some embodiments of this application. In some embodiments, such as Figure 5 As shown, this application embodiment also provides an electronic device 500, including a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor 501. When the program is executed by the processor 501, it implements the various processes of the above-described palm image quality evaluation model training method embodiment, or implements the various processes of the above-described palm image quality evaluation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0112] It should be noted that the electronic devices in the embodiments of this application include the mobile electronic devices and non-mobile electronic devices described above.

[0113] This application also provides a non-transitory computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the above-described palm image quality evaluation model training method embodiment, or implements the various processes of the above-described palm image quality evaluation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0114] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0115] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the various processes of the above-described palm image quality evaluation model training method embodiment, or implements the various processes of the above-described palm image quality evaluation method embodiment.

[0116] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0117] This application embodiment also provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor. The processor is used to run programs or instructions to implement the various processes of the above-described palm image quality evaluation model training method embodiment, or to implement the various processes of the above-described palm image quality evaluation method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0118] It should be understood that the chip mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.

[0119] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, and displayed data), and signals (including but not limited to signals transmitted between the user terminal and other devices), such as palm image samples and palm images, involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0120] For example, the aforementioned palm image quality assessment model training device or palm image quality assessment device can be used in financial institutions, public transportation, and educational settings. Financial institutions include, but are not limited to, banks, securities firms, and insurance companies; public transportation includes, but is not limited to, subway stations, high-speed rail stations, and airports; and educational settings include examination rooms and classroom attendance systems. In all these scenarios, full authorization has been obtained from all parties before data is acquired through the palm image quality assessment model training device or palm image quality assessment device.

[0121] Of course, the palm image quality evaluation model training device or palm image quality evaluation device can also be used in private scenarios, such as company attendance. When used in private scenarios, user authorization must be obtained before data can be obtained through the palm image quality evaluation model training device or palm image quality evaluation device.

[0122] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0123] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the related technology, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0124] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

[0125] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0126] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

Claims

1. A palm image quality evaluation model training method, characterized in that, The method comprises: acquiring a plurality of palm part image samples; classifying the plurality of palm part image samples by a classifier to determine a classification result of each palm part image sample; determining a classification accuracy of each palm part image sample according to the classification results of the plurality of palm part image samples; determining a quality label of a corresponding palm part image sample according to the classification accuracy of each palm part image sample; evaluating the plurality of palm part image samples by a palm part image quality evaluation model to obtain a quality score of each palm part image sample, the quality score representing the image quality of the palm part image sample; constructing a first loss function according to the quality label and the quality score of each palm part image sample, and training the palm part image quality evaluation model based on the first loss function.

2. The palm image quality evaluation model training method according to claim 1, characterized in that, The method further comprises: determining a within-class score of the palm part image sample according to other palm part image samples having the same classification result as the palm part image sample; determining an inter-class score of the palm part image sample according to other palm part image samples having different classification results from the palm part image sample; determining the classification accuracy of the palm part image sample according to the within-class score and the inter-class score.

3. The method of claim 1 or 2, wherein, After training the palm part image quality evaluation model based on the first loss function, the method further comprises: evaluating the plurality of palm part image samples by the palm part image quality evaluation model to obtain a quality score of each palm part image sample; determining a weight of a corresponding palm part image sample according to the quality score of each palm part image sample; constructing a second loss function according to the weight of each palm part image sample, and training the classifier based on the second loss function; iteratively performing the step of classifying the plurality of palm part image samples by the classifier to determine a classification result of each palm part image sample until a preset stopping condition is met.

4. The method of claim 3, wherein, The method further comprises: determining a stable factor; for each palm part image sample, determining a weight of the palm part image sample based on the quality score of the palm part image sample and the stable factor; wherein the weight of the palm part image sample is negatively correlated with the quality score.

5. The method of claim 3, wherein the method further comprises: The method further comprises: for each palm part image sample, calculating a cross-entropy of the classification result of the palm part image sample and a classification result label; constructing the second loss function by combining the cross-entropy corresponding to each palm part image sample and the weight of each palm part image sample.

6. The method of claim 3, wherein the method further comprises: The method further comprises: if the number of iterations reaches a preset number threshold, it is determined that the preset stopping condition is met; if the classification accuracy obtained by verifying the classifier by a plurality of verification palm part image samples converges, it is determined that the preset stopping condition is met; if the quality score obtained by verifying the palm part image quality evaluation model by the plurality of verification palm part image samples converges, it is determined that the preset stopping condition is met.

7. The palm image quality assessment model training method according to claim 1, characterized in that, The quality label of each palm image sample is determined according to the classification accuracy of each palm image sample. If the classification accuracy of each palm image sample is not less than a preset accuracy threshold, the quality label of the palm image sample is determined as a positive sample. If the classification accuracy of each palm image sample is less than a preset accuracy threshold, the quality label of the palm image sample is determined as a negative sample.

8. A method for evaluating the quality of a palm image, characterized by The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. 9.A device for training a palm image quality assessment model, characterized in that, The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7.

10. A hand part image quality evaluation apparatus characterized by comprising: The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in any one of claims 1 to 7. The palm image quality evaluation model is trained based on the palm image quality evaluation model training method in

Images