CN114863178A - Image data input detection method and system for neural network vision system - Google Patents

Abstract

The invention discloses an image data input detection method and system oriented to a neural network vision system. Given a neural network model and its training image data set, the training image data set is input to the neural network model and intermediate results are collected to obtain the neural network hidden image data set. Contains features; uses a Gaussian mixture model to fit the intermediate results, obtains model parameters, and collects the path frequency of the training image data set to calculate the probability; input the image data to be tested into the neural network model, and collect the intermediate results according to the method in step 1 ; Use the Gaussian mixture model in step 2 to calculate the generation probability and inter-layer transition probability of the intermediate result, and use the joint probability estimation model to perform fast probability estimation to verify whether the input image data to be tested is valid.

Description

Image data input detection method and system for neural network vision system

technical field

The invention relates to an image data input detection method and system for a neural network vision system, belonging to the technical fields of deep neural network model input verification, intelligent software quality assurance, image data processing and the like.

Background technique

Software systems based on deep neural networks have been widely used in various fields of production and life in recent years, bringing great convenience to people's lives. An important field of application of deep learning models is machine vision, which also includes sub-tasks such as image classification, object detection, and semantic segmentation. Although software systems based on deep neural network technology have achieved good performance in these tasks, neural network models often do not achieve 100% accuracy. Moreover, the prediction process of the deep neural network model is difficult to explain, so whether the prediction result of the visual model is correct cannot be judged by a simple method.

Vision systems based on neural network models have been used in areas with high security requirements such as face recognition and video surveillance, and wrong prediction results may have serious consequences. In order to ensure the safety of machine vision systems based on deep neural networks, some abnormal input detection techniques for image data have gradually emerged. This kind of method can estimate the probability of abnormal prediction of neural network vision system for a given image data sample. If the estimated abnormal prediction probability is greater than a predetermined threshold, the sample will be rejected and a warning will be given to ensure the reliable operation of the machine vision intelligent system.

There are still some limitations in the applicability and reliability of the existing image input detection technology. Some methods discriminative are based on discriminative models, which model the input image detection problem as a binary classification problem of normal data and abnormal data. The construction process of these methods requires the participation of some abnormal input images, and obtaining and labeling image samples increases the cost of the application. In addition, due to the participation of abnormal data, discriminative methods may overfit to specific data distributions, and it is difficult to guarantee the same generalization performance for image data from different distributions. Others have model-independent characteristics. These methods use specific strategies to retrain the deep neural network model, which not only increases the cost and difficulty of applying detection technology, but also the retrained model often leads to performance degradation, which in many application scenarios are unacceptable.

SUMMARY OF THE INVENTION

Purpose of the invention: Aiming at the problems and deficiencies in the prior art, the present invention provides an image data input detection method and system based on joint modeling of the intermediate layer for the neural network vision system. In the machine vision intelligent system based on the deep neural network, the intermediate results generated in the inference process of the deep neural network system are collected and analyzed, and the input image sample data that cannot be correctly predicted by the deep neural network model is detected.

The advantages of the present invention are as follows: 1. The use scene is not strictly limited by the complexity of the neural network model and the use scene, especially in some scenes with abnormal image data that are difficult to collect. Second, the method has high accuracy for the validity detection of input image data, and can effectively discriminate between valid and invalid input images, thereby avoiding meaningless prediction results. Third, the method requires less time to detect input images, can meet the requirements of real-time verification of correctness, and can be deployed in running machine vision models for input detection.

Technical solution: an image data input detection method oriented to a neural network vision system, comprising the following steps:

Step 1: Neural network hidden feature extraction. Given a neural network model and its training image dataset, input the training image dataset to the neural network model and collect intermediate results to obtain the hidden features of the neural network.

First, according to the given neural network model to be tested, rewrite its forward propagation sub-process to derive intermediate results in the inference process; then use the given training image dataset to input the image dataset into the neural network model and collect the intermediate layer hidden features.

Step 2: Express joint probability modeling in space. Use a Gaussian mixture model to fit the intermediate results, obtain the model parameters, and collect the path frequencies of the training image dataset to calculate the probability.

Constructing a joint probability model requires two steps. First, a generative model based on probability graph is established by using the hidden features of the intermediate layer generated in step 1. Then map the intermediate layer features into discrete space to obtain the transition probability of image data samples in adjacent intermediate layers;

Step 3: Joint probability estimation model. Input the image data to be tested into the neural network model, and collect the intermediate results according to the method as in step 1. Use the Gaussian mixture model in step 2 to calculate the generation probability and inter-layer transition probability of the intermediate results, and use the joint probability estimation model to perform fast probability estimation to verify whether the input image data to be tested is valid.

The neural network is a deep neural network, which refers to a machine learning model for image data feature extraction and prediction formed by using neurons to perform hierarchical connections, including an input layer, a hidden layer and an output layer. The layer contains a large number of neurons, and the neurons are connected between the layers. The input layer transfers the image data to the output layer, and outputs the prediction result, such as the category to which the image data belongs; the middle layer result is the neural network in the input layer and the output layer. The hidden feature data output by the neurons of the hidden layer between the two; the neuron is the structure that uses the built-in function to operate the input data for the neuron input, and outputs the result of the operation; the input image data is Refers to a single or batch of image data samples that conform to the input format of a deep neural network model.

In the step 1, the neural network model to be tested refers to the pre-training that contains complete model information (such as model architecture, neuron parameters, etc.) and has complete operation authority (such as collecting model intermediate results, modifying model parameters, etc.). Model.

In the step 1, the input of the training image data set to the neural network model and the collection of intermediate results means that the forward propagation process covering the neural network enables the neural network to retain the hidden features of the output of the neurons in the middle layer, and collect training images. The mid-level hidden features of the dataset. Because the neurons in the neural network model are arranged in layers, the intermediate results can also be divided into multiple layers. In theory, all the hidden features of the middle layer can be collected and utilized. However, due to actual hardware limitations and the results obtained by sampling have a high accuracy rate, it is not necessary to use all the intermediate results. In actual use, a uniform extraction strategy or an extraction strategy according to the model structure can be used to sample the intermediate results. In practical applications, if the amount of feature data in the middle layer is still large, redundant information can be removed by dimensionality reduction. If global pooling can achieve good results, the PCA dimension reduction algorithm can also be used together with the present invention.

In the second step, the Gaussian mixture model refers to a multi-dimensional Gaussian mixture clustering algorithm, which has a fitting effect on the multi-dimensional complex data distribution. The common expectation-maximization (EM) algorithm can be used to obtain the approximate solution of the Gaussian mixture model, and several Gaussian components and corresponding weights can be obtained, and then the sample probability can be obtained by the generative method. These model parameters are the building blocks of a subsequent framework for evaluating the probability of new samples.

In the second step, the joint probability modeling of the representation space refers to the joint probability modeling of the implicit features in the representation space of the deep neural network, and the purpose is to model the reasoning process of the neural network. The representation space refers to the data space of the hidden features of the intermediate layer, including the intermediate representation features of the input data of the model.

In the second step, the fitting of the hidden features of the middle layer is to use a Gaussian mixture model (GMM) to establish a probability distribution model of the hidden features of the middle layer. A Gaussian mixture model is built on the output of the ith intermediate layer using the expectation-maximization (EM) algorithm, and the parameters Θ _i of the K _i Gaussian components and the weight Φ _i of each Gaussian component are obtained. The parameters Φ _i and Θ _i in the probabilistic graphical model are the basic materials for evaluating the anomaly probability of image data.

In the second step, the _probability is _calculated by ^collecting the ^path frequency of the training image data set. m is the training set size. The probability p(z _i |z _{i -1} ) of the transition from zi _-1 to zi is calculated, which provides basic parameters for the subsequent evaluation of the sample probability.

In the second step, the transition probability of the adjacent intermediate layer refers to the transition probability between the adjacent layers of the discrete component _zi corresponding to the intermediate layer feature x _i on the training image data set; the discrete component z _i means that according to the clustering result of the GMM of the i-th layer, the input image data belongs to the cluster of the intermediate-layer feature x _i of the i-th layer as z _i . Further, calculate the probability P(z _i |z _{i -1} ) of the transition from zi _-1 to zi on the training image data set, that is, the transition probability from the i-1th layer to the i layer. The possible values of zi are K _i , and the transition probability from the _i -1th layer to the i layer is a matrix of size K _i-1 ×K _i .

In the third step, the joint probability estimation model refers to estimating the joint probability of the output sequence {x ₁ ,x ₂ ,...,x _m } of the middle layer through the probability graph model. The time complexity of directly calculating the joint probability P(x ₁ , x ₂ , . . . , x _m ) increases exponentially with respect to m, so the present invention uses a fast forward algorithm based on dynamic programming. make

α _i (z _i )≡P(z _i ,x ₁ ,x ₂ ,…,x _i ),

Then α _i ( ) can be generated by the following recursive process:

where K _i-1 represents the number of Gaussian components in the i-1th layer, P(x _i |z _i ) is given by the GMM of the i-th layer, and P(z _i |z _i-1 ) is the transition probability. Thus we get the joint probability of the output of the middle layer with linear time complexity.

In the third step, the verification of whether the input image data to be tested is valid refers to judging whether the joint probability of the image data is greater than a preset threshold; the threshold is a value between 0 and 1, and the closer it is to 0, the more inclined it is. For the accuracy of the method, the closer it is to 1, the more likely the recall rate of the method is; a feasible strategy in practice is to set a threshold that makes most image data (such as 90%) normal according to the joint probability of the training image data set.

An image data input detection system oriented to a neural network vision system, comprising:

Neural network latent feature extractor: Given a neural network model and its training image dataset, input the training image dataset to the neural network model and collect intermediate results to obtain neural network latent features.

First, according to the given neural network model to be tested, rewrite its forward propagation sub-process to derive intermediate results in the inference process; then use the given training image dataset to input the image dataset into the neural network model and collect the intermediate layer hidden features.

Representation Space Joint Probabilistic Modeling Tool: Fits intermediate results using a Gaussian mixture model, obtains model parameters, and collects training image dataset path frequencies to calculate probabilities.

Constructing a joint probability model requires two steps. First, a generative model based on probability graphs is built using the intermediate layer latent features generated by the neural network latent feature extractor. Then map the intermediate layer features into discrete space to obtain the transition probability of image data samples in adjacent intermediate layers;

Joint probability estimation model; input the image data to be tested into the neural network model, and use the neural network latent feature extractor to collect its intermediate results. Use the Gaussian mixture model in the representation space joint probability modeling tool to calculate the generation probability and interlayer transition probability of the intermediate results, and use the joint probability estimation model to perform fast probability estimation to verify whether the input image data to be tested is valid.

A computer device comprising a memory, a processor and a computer program stored in the memory and running on the processor, the processor implements the above-mentioned image data input oriented to a neural network vision system when the processor executes the above-mentioned computer program Detection method.

A computer-readable storage medium storing a computer program for executing the above-mentioned image data input detection method for a neural network vision system.

Beneficial effects: The present invention can make up for the deficiency that the ordinary neural network visual model is difficult to identify abnormal image data, thereby ensuring the safety of the intelligent system. Compared with the existing neural network model input data verification technology, the present invention does not need to use abnormal data for training, and is not prone to overfitting to specific data. The present invention can efficiently detect and evaluate the validity of the input image data, and utilize the validity of the evaluation, thereby realizing the real-time screening of the test images and improving the effect of the actual deployment of the neural vision system.

Description of drawings

1 is a schematic diagram of a method according to an embodiment of the present invention;

2 is a schematic diagram of a neural network implicit feature extractor according to an embodiment of the present invention;

3 is a flowchart representing a spatial joint probability modeling tool according to an embodiment of the present invention;

4 is a functional schematic diagram of a joint probability estimation model according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of a probability graph model according to an embodiment of the present invention.

Detailed ways

Below in conjunction with specific embodiments, the present invention will be further illustrated, and it should be understood that these embodiments are only used to illustrate the present invention and not to limit the scope of the present invention. The modifications all fall within the scope defined by the appended claims of this application.

As shown in Figure 1, the image data input detection method for neural network vision system consists of two stages: training and deployment. In the training phase, the latent feature extractor is first used to input the training image data into the neural network model to generate the latent features generated by each intermediate layer. Thereafter, a Gaussian mixture model is trained on the intermediate layer features using a representation space joint modeling tool, and using the clustering results of the Gaussian mixture model, transition probabilities between adjacent layers are calculated. In the deployment stage, for any image data to be tested, it is firstly input into the neural network to obtain intermediate layer features, and then the joint probability estimation model is used to calculate the joint probability of the input intermediate layer features. Finally, the joint probability value is compared with a predetermined threshold to decide whether to accept or reject the image data to be tested. The whole framework consists of three modules corresponding to three steps: neural network latent feature extractor, representation space joint probability modeling tool, and joint probability estimation model.

Image data input detection system for neural network vision system, including:

Step 1: Given a neural network model and its training image data set, input the training image data set to the neural network model and collect intermediate results to obtain the hidden features of the neural network.

As shown in Figure 2, the forward propagation process of the overlay neural network visual model makes it retain the hidden features of the output of the neurons in the middle layer, input the training image dataset into the neural network, and extract the hidden features generated by the middle layers of the model {x ₁ ,x ₂ ,…,x _m }, where m is the number of intermediate layers. In theory, all the intermediate features of these image data can be collected and utilized, but due to actual hardware limitations, and the results obtained by sampling have a very high accuracy, it is not necessary to use all the intermediate results. In actual use, a uniform extraction strategy or an extraction strategy according to the model structure can be used for sampling. In practical applications, if the amount of feature data in the middle layer is still large, the redundant information output by the middle layer of the image data can be removed by dimensionality reduction. If global pooling can achieve good results, the PCA dimension reduction algorithm can also be used together with the present invention. Save the processed intermediate layer results for subsequent training.

In this step, it is required that the neural network vision system to be tested contains complete model information (such as model architecture, neuron parameters, etc.) and has complete operation authority (such as collecting model intermediate results, modifying model parameters, etc.). Not suitable for a black-box model that only provides API services.

Step 2: Use the representation space joint probability modeling to model the reasoning process of the neural network visual model to the training set image data.

As shown in Figure 3, the Gaussian mixture model is trained layer by layer on the intermediate layer data obtained in step 1. The Gaussian mixture model of each layer can reasonably select parameters, such as the number of Gaussian components, according to the size of the image data and the computing power of the device. Retain the Gaussian mixture model on the data of each intermediate layer, and perform the clustering operation on the image data of the training set to obtain the output clustering result _zi of the image data intermediate layer, and calculate the transition probability P between adjacent intermediate layer clusters according to the result. (z _i |z _i-1 ), save it for use in online image data to be measured.

Representation space joint probability modeling refers to the joint probability modeling of implicit features in the representation space of deep neural network, and its purpose is to model the reasoning process of neural vision network on image data. The representation space refers to the data space of the hidden features of the middle layer, including the intermediate representation features of the model input data. This method uses the probabilistic graph model shown in Figure 5 to represent the joint probability of the intermediate layers. The intermediate layers of the neural network are in a corresponding relationship with the layers in the probabilistic graphical model, where x _i is the output of the ith intermediate layer of the neural network model, and z _i is the hidden state of the _i -th layer of the probabilistic graphical model, and Φi and _Θi are the parameters of the method.

Fitting the hidden features of the middle layer separately is to use the Gaussian mixture model (GMM) to establish the probability distribution model of the hidden features of the middle layer. A Gaussian mixture model is built on the output of the ith intermediate layer using the expectation-maximization (EM) algorithm, and the parameters Θ _i of the K _i Gaussian components and the weight Φ _i of each Gaussian component are obtained. The parameters Φ _i and Θ _i in the probabilistic graphical model are the basic materials for evaluating the anomaly probability of image data.

The transition probability of adjacent intermediate layers refers to the transition probability of the discrete component _zi corresponding to the intermediate layer feature x _i between adjacent layers on the training image dataset; the discrete component _zi refers to the The clustering result, the input data in the i-th intermediate layer feature _xi belongs to the cluster _zi . Calculate the probability P(z _i |z _{i -1} ) of the transition from zi _-1 to zi on the training data set, that is, the transition probability from layer i-1 to layer i. The possible values of zi are K _i , and the transition probability from the _i -1th layer to the i layer is a matrix of size K _i-1 ×K _i .

Step 3: Use the joint probability estimation model to analyze the validity of the image data to be tested and report.

As shown in Figure 4, one or a batch of image data to be tested is input into the neural network model, the inference process is performed and the intermediate results are extracted, and the intermediate layer results are dimensionally reduced and saved. Note that the middle layer of the model selected in this step should be consistent with the operation of step 1, and the dimensionality reduction operation should also be consistent. Input the output data of the middle layer into the Gaussian mixture model corresponding to the middle layer _, and obtain the probability P ₍ x _i | The joint probability of the intermediate layer output sequence {x ₁ ,x ₂ ,…,x _m } is estimated by a probabilistic graphical model. The time complexity of directly computing the joint probability P(x ₁ ,x ₂ ,...,x _m ) grows exponentially with respect to m, so a fast forward algorithm based on dynamic programming is used. make

α _i (z _i )≡P(z _i ,x ₁ ,x ₂ ,…,x _i ),

Then α _i ( ) can be generated by the following recursive process:

Finally, the joint probability of the image data intermediate layer feature sequence {x ₁ ,x ₂ ,...,x _m } is obtained. Determine whether the joint probability of the image data to be tested is greater than a preset threshold; the threshold is a value between 0 and 1. The closer to 0, the more accurate the method is, and the closer to 1, the more likely to be the recall rate of the method; practice One strategy that works in , is to set a threshold that makes the majority of image samples (eg, 90%) normal, based on the joint probability of the training image dataset.

An image data input detection system oriented to a neural network vision system, comprising:

Neural network latent feature extractor: Given a neural network model and its training image dataset, input the training image dataset to the neural network model and collect intermediate results to obtain neural network latent features.

First, according to the given neural network model to be tested, rewrite its forward propagation sub-process to derive intermediate results in the inference process; then use the given training image dataset to input the image dataset into the neural network model and collect the intermediate layer hidden features.

Representation Space Joint Probabilistic Modeling Tool: Fits intermediate results using a Gaussian mixture model, obtains model parameters, and collects training image dataset path frequencies to calculate probabilities.

Constructing a joint probability model requires two steps. First, a generative model based on probability graphs is built using the intermediate layer latent features generated by the neural network latent feature extractor. Then map the intermediate layer features into discrete space to obtain the transition probability of image data samples in adjacent intermediate layers;

Joint probability estimation model; input the image data to be tested into the neural network model, and use the neural network latent feature extractor to collect its intermediate results. Use the Gaussian mixture model in the representation space joint probability modeling tool to calculate the generation probability and interlayer transition probability of the intermediate results, and use the joint probability estimation model to perform fast probability estimation to verify whether the input image data to be tested is valid.

Obviously, those skilled in the art should understand that the steps of the image data input detection method oriented to the neural network vision system or the modules of the image data input detection system oriented to the neural network vision system according to the above-mentioned embodiments of the present invention can use a general-purpose computing device They can be centralized on a single computing device or distributed on a network composed of multiple computing devices, optionally, they can be implemented with program codes executable by the computing devices, so that they can be stored in The storage device is performed by the computing device, and in some cases, the steps shown or described may be performed in an order different from that herein, or separately fabricated into individual integrated circuit modules, or multiple of them. Each module or step is fabricated into a single integrated circuit module to implement. As such, embodiments of the present invention are not limited to any particular combination of hardware and software.

Claims (10)

1. An image data input detection method for a neural network vision system is characterized by comprising the following steps:

the method comprises the following steps: extracting implicit characteristics of a neural network; giving a neural network model and a training image data set thereof, transmitting the training image data set to the neural network model and collecting an intermediate result to obtain implicit characteristics of the neural network;

step two: representing spatial joint probability modeling; fitting the intermediate result by using a Gaussian mixture model to obtain model parameters, and collecting training image data set path frequency calculation probability;

step three: a joint probability estimation model; inputting image data to be detected into a neural network model, and collecting an intermediate result according to the method in the step one; and calculating the generation probability and the interlayer transition probability of the intermediate result by using the Gaussian mixture model in the step two, performing rapid probability estimation by using a joint probability estimation model, and verifying whether the input image data to be detected is effective or not.

2. The image data input detection method for the neural network visual system as claimed in claim 1, wherein the neural network is a deep neural network, which is a machine learning model for image data feature extraction and prediction formed by hierarchical connection of neurons, and comprises an input layer, a hidden layer and an output layer; the layers comprise neurons, the neurons are connected among the layers, image data are transmitted from the input layer to the output layer, and a prediction result is output; the intermediate layer result is implicit characteristic data output by neurons of an implicit layer of the neural network between the input layer and the output layer; the neuron is a structure that performs an operation on input data by using a built-in function or the like for neuron input and outputs an operation result; the input image data refers to a single or a batch of image data samples conforming to the input format of the deep neural network model.

3. The method for detecting image data input facing to the neural network vision system as claimed in claim 1, wherein in the first step, according to the given neural network model to be detected, the forward propagation subprocess of the neural network model is rewritten so that the neural network model can derive the intermediate result in the inference process; a given training image dataset is then used, which is input into the neural network model and the intermediate layer implicit features are collected.

4. The neural network vision system-oriented image data input detection method as claimed in claim 1, wherein the intermediate results are sampled by using a uniform extraction strategy or a model structure extraction strategy.

5. The method for detecting image data input facing a neural network vision system as claimed in claim 1, wherein in the second step, two steps are required for constructing the joint probability model, and firstly, a generative model based on a probability map is established by using the intermediate layer implicit characteristics generated in the first step; then mapping the characteristics of the middle layer into a discrete space to obtain the transition probability of the image data sample in the adjacent middle layer;

the representation space refers to a data space with hidden characteristics of the middle layer and comprises middle representation characteristics of model input data;

the step of respectively fitting the intermediate layer hidden features is to establish a probability distribution model of the intermediate layer hidden features by using a Gaussian mixture model; using expectation-maximization algorithm to build a Gaussian mixture model on the ith intermediate layer output to obtain K _i Parameters theta of the respective Gaussian components _i And the weight of each Gaussian component _i (ii) a Parameter phi in probabilistic graphical model _i And Θ _i Is a basic material for evaluating the abnormal probability of the image data.

6. The neural-network visual system-oriented image data input detection method according to claim 1, wherein in the second step, the transition probabilities of adjacent intermediate layers refer to the intermediate layer feature x on the training image data set _i Corresponding discrete component z _i Transition probability between adjacent layers; the discrete component z _i Means that according to the clustering result of the GMM of the ith layer, the input image data has the characteristic x of the middle layer of the ith layer _i The cluster is z _i (ii) a Computing z on a training image dataset _i-1 To z _i Probability of transition P (z) _i |z _i-1 ) I.e., transition probability from layer i-1 to layer i; z is a radical of _i Possible values are K _i The transition probability from layer i-1 to layer i is of size K _i-1 ×K _i Of the matrix of (a).

7. The method for detecting image data input facing neural network vision system as claimed in claim 1, wherein in step three, said joint probability estimation model refers to estimating middle layer output sequence { x } through probability map model ₁ ,x ₂ ,…,x _m J, directly calculating the joint probability P (x) ₁ ,x ₂ ,…,x _m ) The time complexity of (a) increases exponentially with respect to (m), and therefore a fast forward algorithm based on dynamic programming is used; order to

α _i (z _i )≡P(z _i ,x ₁ ,x ₂ ,…,x _i )，

Then alpha _i (. h) is generated by the following recursive process:

wherein K _i-1 Indicates the number of Gaussian components of the i-1 st layer, P (x) _i |z _i ) Given by the GMM of layer i, P (z) _i |z _i-1 ) Is the transition probability.

8. An image data input detection system for a neural network vision system, comprising:

the neural network implicit feature extractor: giving a neural network model and a training image data set thereof, transmitting the training image data set to the neural network model and collecting an intermediate result to obtain implicit characteristics of the neural network;

firstly, rewriting a forward propagation subprocess of a given neural network model to be tested according to the neural network model to be tested so that an intermediate result is derived in a reasoning process; then, inputting the image data set into a neural network model by using a given training image data set and collecting implicit characteristics of the middle layer;

representing a spatial joint probability modeling tool: fitting the intermediate result by using a Gaussian mixture model to obtain model parameters, and collecting training image data set path frequency calculation probability;

two steps are needed for constructing the joint probability model, and firstly, a generative model based on a probability graph is established by using intermediate layer hidden features generated by a neural network hidden feature extractor. Then mapping the characteristics of the middle layer into a discrete space to obtain the transition probability of the image data in the adjacent middle layer;

a joint probability estimation model; inputting image data to be detected into a neural network model, and collecting intermediate results according to a neural network implicit feature extractor; and calculating the generation probability and the interlayer transition probability of the intermediate result by using a Gaussian mixture model in the space joint probability modeling tool, performing rapid probability estimation by using a joint probability estimation model, and verifying whether the input image data to be detected is effective or not.

9. A computer device, characterized by: the computer device comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the computer program to realize the image data input detection method for the neural network vision system as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium characterized by: the computer-readable storage medium stores a computer program for executing the image data input detection method for a neural network vision system according to any one of claims 1 to 7.

Images