CN113780444B - Training method of tongue fur image classification model based on progressive learning - Google Patents
Training method of tongue fur image classification model based on progressive learning Download PDFInfo
- Publication number
- CN113780444B CN113780444B CN202111086464.5A CN202111086464A CN113780444B CN 113780444 B CN113780444 B CN 113780444B CN 202111086464 A CN202111086464 A CN 202111086464A CN 113780444 B CN113780444 B CN 113780444B
- Authority
- CN
- China
- Prior art keywords
- target
- resolution
- regularized
- classification model
- tongue
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a training method of a tongue fur image classification model based on progressive learning, which comprises the following steps: acquiring an original training data set at least comprising a plurality of tongue fur images; determining a target resolution set according to a preset progressive learning resolution amplification mode; determining a regularized intensity value set according to a preset regularized intensity amplification mode of progressive learning; performing a preset resolution resetting operation on each tongue fur image in the original training data set to obtain a target training data set; and training the preset tongue fur image classification model for multiple times based on the target training data set and the regularized intensity value set. Therefore, the invention can train the tongue coating images with different resolutions by using different regularization intensities, thereby accelerating the training speed of the tongue coating image classification model and improving the classification accuracy of the tongue coating image classification model after training. The invention can be applied to digital medical systems.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a training method, a training device, computer equipment and a storage medium of a tongue fur image classification model based on progressive learning.
Background
With the rapid development of deep learning and artificial intelligence, it has become practical to assist a user in classifying tongue fur images by using a tongue fur image classification algorithm based on a convolutional neural network. According to the technical characteristics of the artificial neural network, the classification accuracy of the tongue fur image classification model can be influenced by the training process of the tongue fur image classification model to a certain extent. In the existing research, in order to improve the classification accuracy of the tongue coating image classification model, people try to improve the training effect of the tongue coating image classification model by increasing the resolution of the tongue coating image in the training data, so as to improve the classification accuracy of the tongue coating image classification model. However, it has been found in practice that this method is liable to cause a decrease in classification accuracy after the resolution of the tongue coating image is increased to a certain extent. Therefore, how to train the tongue coating image classification model scientifically and effectively to improve the classification accuracy of the tongue coating image classification model is a technical problem to be solved urgently.
Disclosure of Invention
The invention aims to solve the technical problem that in the current training method of the tongue fur image classification model, the classification accuracy of the tongue fur image classification model is lower.
In order to solve the technical problems, the first aspect of the invention discloses a training method of a tongue fur image classification model based on progressive learning, which comprises the following steps:
acquiring a target video to be processed;
performing dynamic target detection on the target video based on a preset dynamic target detection model to obtain a dynamic target region, a non-dynamic target region and depth information corresponding to the dynamic target region in the target video;
determining an image insertion area of the target video according to the dynamic target area and the non-dynamic target area;
determining target depth information corresponding to an image insertion area of the target video according to the depth information corresponding to the dynamic target area;
and inserting the predetermined advertisement image into a position corresponding to the target depth information in an image insertion area of the target video so as to finish the processing of the target video.
The invention discloses a training device of a tongue fur image classification model based on progressive learning, which comprises the following components:
the acquisition module is used for acquiring a target video to be processed;
the detection module is used for carrying out dynamic target detection on the target video based on a preset dynamic target detection model to obtain a dynamic target area, a non-dynamic target area and depth information corresponding to the dynamic target area in the target video;
the determining module is used for determining an image insertion area of the target video according to the dynamic target area and the non-dynamic target area;
the determining module is further configured to determine target depth information corresponding to an image insertion area of the target video according to the depth information corresponding to the dynamic target area;
and the inserting module is used for inserting the predetermined advertisement image into the position corresponding to the target depth information in the image inserting area of the target video so as to finish the processing of the target video.
A third aspect of the invention discloses a computer device comprising:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program code stored in the memory to perform some or all of the steps in the training method for a tongue fur image classification model based on progressive learning disclosed in the first aspect of the present invention.
A fourth aspect of the invention discloses a computer storage medium storing computer instructions for performing part or all of the steps of the training method of the tongue fur image classification model based on progressive learning disclosed in the first aspect of the invention when the computer instructions are invoked.
In the embodiment of the invention, an original training data set comprising a plurality of tongue fur images is obtained, a target resolution set is determined according to a preset progressive learning resolution amplification mode, a regularized intensity value set corresponding to the target resolution set is determined according to a preset progressive learning regularized intensity amplification mode, a preset resolution resetting operation is carried out on each tongue fur image in the original training data set to obtain a target training data set, and finally, a preset tongue fur image classification model is trained for a plurality of times based on the target training data set and the regularized intensity value set, so that a target loss function of the tongue fur image classification model is converged, and training of the tongue fur image classification model is completed. The method comprises the steps of determining a target resolution set and a regularized intensity value set corresponding to the target resolution set in advance, resetting the resolutions of a plurality of tongue fur images in an original training data set to be the target resolutions in the target resolution set to form a target training data set comprising tongue fur images with various resolutions, so that when a tongue fur image classification model is trained, training models can be carried out on tongue fur images with different resolutions by using different regularized intensities, the training speed of the tongue fur image classification model can be increased, and the classification accuracy of the tongue fur image classification model after training is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a training method of a tongue fur image classification model based on progressive learning according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a training device of a tongue fur image classification model based on progressive learning according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a computer device according to an embodiment of the present invention;
fig. 4 is a schematic structural view of a computer storage medium according to an embodiment of the present invention.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The terms first, second and the like in the description and in the claims and in the above-described figures are used for distinguishing between different objects and not necessarily for describing a sequential or chronological order. Furthermore, the terms "comprise" and "have," as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, apparatus, article, or article that comprises a list of steps or elements is not limited to only those listed but may optionally include other steps or elements not listed or inherent to such process, method, article, or article.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the invention. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly appreciate that the embodiments described herein may be combined with other embodiments.
The invention discloses a training method, a device, computer equipment and a storage medium of a tongue coating image classification model based on progressive learning, which are used for acquiring an original training data set comprising a plurality of tongue coating images, determining a target resolution set according to a preset progressive learning resolution amplification mode, determining a regularization intensity value set corresponding to the target resolution set according to a preset progressive learning regularization intensity amplification mode, executing preset resolution resetting operation on each tongue coating image in the original training data set to obtain a target training data set, and finally training the preset tongue coating image classification model for multiple times based on the target training data set and the regularization intensity value set to enable a target loss function of the tongue coating image classification model to converge so as to complete training of the tongue coating image classification model. The method comprises the steps of determining a target resolution set and a regularized intensity value set corresponding to the target resolution set in advance, resetting the resolutions of a plurality of tongue fur images in an original training data set to be the target resolutions in the target resolution set to form a target training data set comprising tongue fur images with various resolutions, so that when a tongue fur image classification model is trained, training models can be carried out on tongue fur images with different resolutions by using different regularized intensities, the training speed of the tongue fur image classification model can be increased, and the classification accuracy of the tongue fur image classification model after training is improved. The following will describe in detail.
Example 1
Referring to fig. 1, fig. 1 is a flowchart of a training method of a tongue fur image classification model based on progressive learning according to an embodiment of the present invention. As shown in fig. 1, the training method of the tongue fur image classification model based on progressive learning may include the following operations:
101. an original training dataset comprising at least a plurality of tongue coating images is acquired.
In the step 101, the original training dataset may include a plurality of images (i.e. tongue coating images) with human tongue coating, and each tongue coating image may have a corresponding labeling label, where the labeling labels are used to classify the tongue coating images into normal tongue coating, peeled coating, little coating and no coating, and the four types are all used. The labeling of each tongue coating image may be manually set. Alternatively, the tongue coating images in the original training data set may be acquired on a digital medical system of a hospital, and a tongue coating image library may be provided in the digital medical system, for storing tongue coating images uploaded by respective image acquisition terminals in the digital medical system.
102. And determining a target resolution set according to a preset progressive learning resolution amplification mode.
In step 102, the progressive learning strategy is a model training strategy for training the tongue coating image classification model using tongue coating images with progressively higher resolution. Specifically, the tongue coating image classification model is trained one by one, and the resolution of the tongue coating image used in each training is gradually increased. For this reason, it is an important question in each training of the tongue coating image classification model, what resolution of tongue coating image is used for training. The resolution used for each training (i.e., the target resolution) may be pre-calculated and then the target resolutions may be grouped into a target resolution set. The specific calculation method of the target resolution will be described later.
103. Determining a regularized intensity value set according to a preset regularized intensity amplification mode of progressive learning, wherein regularized intensity values in the regularized intensity value set are in one-to-one correspondence with target resolutions in the target resolution set, and regularized intensity values corresponding to the target resolutions are increased along with the increase of the target resolutions.
After each target resolution is calculated in step 103 described above, the regularized intensity values that increase with each target resolution may continue to be calculated to form a regularized intensity value set. The specific calculation method of the regularized intensity value will be described in detail later.
104. And executing preset resolution resetting operation on each tongue fur image in the original training data set to obtain a target training data set, wherein the target training data set in the target training data set corresponds to the target resolution in the target resolution set one by one, and the resolution of the tongue fur image in each group of target training data set is the target resolution corresponding to the group of target training data set.
In step 104, after calculating each target resolution, the resolutions of all tongue fur images in the original training dataset may be reset according to each target resolution, so as to obtain a target training dataset corresponding to each target resolution, thereby forming a target training dataset set. For example, the target resolutions include 256×256, 512×512, and 1024×1024, and after the resolution is reset, a target training dataset with the resolution of the tongue coating image of 256×256, a target training dataset with the resolution of the tongue coating image of 512×512, and a target training dataset with the resolution of the tongue coating image of 1024×1024 are obtained, so as to form a target training dataset set.
105. And training the preset tongue coating image classification model for a plurality of times based on the target training data set and the regularized intensity value set, so that the target loss function of the tongue coating image classification model is converged, and training of the tongue coating image classification model is completed.
In step 105, for example, the target resolutions of the target training data sets in the target training data set are 256×256, 512×512, and 1024×1024 in order, and the corresponding regularized intensity values are 0.2, 0.3, and 0.4 in order. At this time, the tongue coating image classification model may be trained for the first time by using the target training data set with the target resolution of 256×256 and the regularized intensity value of 0.2, the tongue coating image classification model may be trained for the second time by using the target training data set with the target resolution of 512×512 and the regularized intensity value of 0.3, and the tongue coating image classification model may be trained for the third time by using the target training data set with the target resolution of 1024×1024 and the regularized intensity value of 0.4, so as to converge the target loss function of the tongue coating image classification model, thereby completing the training of the tongue coating image classification model. Optionally, the tongue coating image classification model adopts a network structure of EfficientNetV 2. In the past studies, many scholars have tried to improve the classification accuracy of the trained tongue coating image classification model by increasing the resolution of the tongue coating image in the training data, but it has been found in practice that after the resolution is improved to some extent, the increase in resolution is liable to cause a decrease in classification accuracy instead. After many practices and analyses, the inventors found that the reason for this was: for the same network structure, the higher the resolution of the input image is, the larger the required network capacity is, and the larger the required regularization strength is, if the resolution of the input image is improved, the regularization strength is still kept unchanged, and the classification effect is likely to be reduced. Therefore, the embodiment of the invention provides a model training mode using different regularization intensities for tongue fur images with different resolutions, and the practice shows that the model training mode can achieve the effects of accelerating the training speed of the tongue fur image classification model and improving the classification accuracy of the tongue fur image classification model after training.
Optionally, the target resolution, regularization strength, image resolution resetting operation and training of the tongue coating image classification model can be completed by a central terminal with strong computing power in the digital medical system, and after the training of the tongue coating image classification model is completed, the trained tongue coating image classification model can be sent to each terminal in the digital medical system by the central terminal to perform the task of image classification.
As can be seen, the training method based on the tongue coating image classification model of progressive learning described in fig. 1 is implemented, an original training dataset including a plurality of tongue coating images is obtained, a target resolution set is determined according to a preset resolution amplification mode of progressive learning, a regularized intensity value set corresponding to the target resolution set is determined according to a preset regularized intensity amplification mode of progressive learning, a preset resolution resetting operation is performed on each tongue coating image in the original training dataset to obtain a target training dataset set, and finally, multiple training is performed on the preset tongue coating image classification model based on the target training dataset set and the regularized intensity value set, so that a target loss function of the tongue coating image classification model converges, and training of the tongue coating image classification model is completed. The method comprises the steps of determining a target resolution set and a regularized intensity value set corresponding to the target resolution set in advance, resetting the resolutions of a plurality of tongue fur images in an original training data set to be the target resolutions in the target resolution set to form a target training data set comprising tongue fur images with various resolutions, so that when a tongue fur image classification model is trained, training models can be carried out on tongue fur images with different resolutions by using different regularized intensities, the training speed of the tongue fur image classification model can be increased, and the classification accuracy of the tongue fur image classification model after training is improved.
In an optional embodiment, the determining the target resolution set according to the preset progressive learning resolution augmentation mode includes:
determining the maximum resolution and the minimum resolution corresponding to the target resolution set;
and determining a plurality of target resolutions from the interval between the maximum resolution and the minimum resolution according to a preset target resolution determination mode, and taking the target resolutions as a target resolution set.
In this alternative embodiment, in calculating the target resolution of the target resolution set, the maximum resolution and the minimum resolution may be determined first, and then a plurality of target resolutions may be selected from among the maximum resolution and the minimum resolution to form the target resolution set.
It will be seen that this alternative embodiment is implemented by first determining a maximum resolution and a minimum resolution, and then selecting a plurality of target resolutions from between the maximum resolution and the minimum resolution to form a target resolution set, thereby enabling calculation of the target resolution.
In an optional embodiment, the determining, according to a preset target resolution determining manner, a plurality of target resolutions from a section between the maximum resolution and the minimum resolution, to be a target resolution set, includes:
determining a plurality of target resolutions from a section between the maximum resolution and the minimum resolution as a target resolution set by the following formula:
wherein S is i Is the ith target resolution in the target resolution set, S 0 Is the minimum resolution, S e Is the maximum resolution, N is the total number of target resolutions in the set of target resolutions.
In this alternative embodiment, N is the total number of target resolutions in the target resolution set, and is also a generation of training the tongue fur image classification modelA number. Each generation of training of tongue coating image classification models requires the use of different target resolution and regularization intensity values. The target resolution used by each generation of training of the tongue fur image classification model can be calculated through the formula. For example, minimum resolution S 0 Set to 256 x 256, maximum resolution S e Set to 512 x 512, the total number of target resolutions (i.e. algebra for training tongue fur image classification model) N in the target resolution set is set to 50, and the calculated target resolution S i 256 x 256, 261.12 x 261.12, 266.24 x 266.24, 271.36 x 271.36, … … in this order
Therefore, by implementing the alternative embodiment, a plurality of target resolutions can be selected from the preset maximum resolution and the preset minimum resolution according to the same interval through the formula, so as to form a target resolution set, so that the target resolutions used in training can be uniformly distributed to the interval between the maximum resolution and the minimum resolution, and the training effect on the tongue fur image classification model is improved.
In an optional embodiment, the determining the regularized intensity value set according to the regularized intensity augmentation mode of the preset progressive learning includes:
determining a maximum regularization intensity value and a minimum regularization intensity value corresponding to the regularization intensity value set;
and determining a plurality of regularized intensity values from the interval between the maximum regularized intensity value and the minimum regularized intensity value according to a preset regularized intensity value determining mode to be used as a regularized intensity value set.
In this alternative embodiment, when calculating the regularized intensity values of the regularized intensity value set, a maximum regularized intensity value and a minimum regularized intensity value may be determined first, and then a plurality of regularized intensity values may be selected from the maximum regularized intensity value and the minimum regularized intensity value to form the regularized intensity value set.
It can be seen that this alternative embodiment is implemented by first determining a maximum regularized intensity value and a minimum regularized intensity value, and then selecting a plurality of regularized intensity values from between the maximum regularized intensity value and the minimum regularized intensity value to form a regularized intensity value set, so that calculation of the regularized intensity values can be achieved.
In an optional embodiment, the determining, according to a preset regularization strength value determining manner, a plurality of regularization strength values from a section between the maximum regularization strength value and the minimum regularization strength value, so as to be a regularization strength value set, includes:
determining a plurality of regularized intensity values from the interval between the maximum regularized intensity value and the minimum regularized intensity value by the following formula to be used as a regularized intensity value set:
wherein D is i Is the ith regularized intensity value in the regularized intensity value set, which corresponds to the ith target resolution in the target resolution set, D 0 Is the minimum regularized intensity value, D e Is the maximum regularized intensity value, N is the total number of regularized intensity values in the regularized intensity value set.
In this alternative embodiment, N is the total number of regularized intensity values in the regularized intensity value set, and is also the algebra for training the tongue coating image classification model. Each generation of training of tongue coating image classification models requires the use of different target resolution and regularization intensity values. The regularization intensity value used by each generation of training of the tongue fur image classification model can be calculated through the formula. For example, a minimum regularized intensity value D 0 Set to 0.1, maximum regularized intensity value D e Set to 0.6, the total number of regularized intensity values (i.e. algebra for training tongue fur image classification model) N in the regularized intensity value set is set to 50, and the regularized intensity value D is calculated i 0.1, 0.11, 0.12, 0.13 and … … in sequence
Therefore, by implementing the optional embodiment, a plurality of regularized intensity values can be selected from the preset maximum regularized intensity value and the preset minimum regularized intensity value according to the same interval through the formula, so that a regularized intensity value set is formed, the regularized intensity values used in training can be uniformly distributed to a section between the maximum regularized intensity value and the minimum regularized intensity value, and the training effect on the tongue fur image classification model is improved.
In an optional embodiment, the original training data set further includes a labeling label corresponding to each of the tongue fur images;
and training the preset tongue coating image classification model for a plurality of times based on the target training data set and the regularized intensity value set, so that a target loss function of the tongue coating image classification model converges, and after training the tongue coating image classification model is completed, the method further comprises:
classifying and predicting each tongue coating image in the original training data set based on the trained tongue coating image classification model to obtain a prediction label and a prediction probability value corresponding to each tongue coating image;
screening noise tongue coating images from all tongue coating images according to the prediction labels, the prediction probability values and the labeling labels corresponding to the tongue coating images, wherein the noise tongue coating images refer to tongue coating images of which the corresponding prediction labels are inconsistent with the labeling labels and the prediction probability values are lower than a preset prediction probability value threshold value;
deleting the noise tongue fur image and the labeling label corresponding to the noise tongue fur image from the original training data set so as to clean the original training data set.
In this alternative embodiment, in practical application, the labeling label corresponding to each tongue coating image in the original training data set may be wrong, for example, the labeling label of the tongue coating image of the normal tongue coating may be set to peel off the tongue coating. If the tongue coating image classification model is trained by using the mislabeled tongue coating images, the classification effect of the tongue coating image classification model is affected. To this end, the original training dataset may be cleaned to remove these falsely labeled tongue coating images from the original training dataset. Specifically, the tongue coating image classification model after training can be used for predicting the tongue coating image in the original training data set, if the prediction label of a certain tongue coating image is inconsistent with the original labeling label and the prediction probability value is lower than the prediction probability value threshold, the labeling label of the tongue coating image can be determined to be wrong, the tongue coating image can be determined to be a noise tongue coating image, and the noise tongue coating image is deleted from the original training data set, so that the original training data set is cleaned. These steps may be repeated further, washing the original training dataset 3-7 times. The cleaned original training data set is used for training the model, so that the classification effect of the tongue fur image classification model can be further improved.
It can be seen that implementing the alternative embodiment, after training the model, the tongue coating image in the original training dataset is predicted by using the trained model to obtain the prediction label and the prediction probability value of the tongue coating image, then the noise tongue coating image is screened out according to the prediction label, the prediction probability value and the original labeling label of the tongue coating image, and finally the noise tongue coating image is deleted from the original training dataset, so as to clean the original training dataset, thereby further improving the classification effect of the tongue coating image classification model.
In an alternative embodiment, the objective loss function of the tongue fur image classification model is:
wherein y is j Labeling label corresponding to jth tongue fur image, p j The predicted probability value corresponding to the j-th tongue coating image is M, and M is the total number of the tongue coating images in the original training data set.
In this alternative embodiment, in the training of the tongue fur image classification model, the model is trained by using the target loss function, which is beneficial to improving the convergence speed of the model and the final classification effect of the model.
Optionally, it is also possible to: and uploading training information of the tongue fur image classification model based on progressive learning to a blockchain.
Specifically, training information of the tongue fur image classification model based on progressive learning is obtained by running the training method of the tongue fur image classification model based on progressive learning, and is used for recording training conditions of the tongue fur image classification model based on progressive learning, such as an obtained original training data set, a target resolution set, a regularized intensity value set, a trained tongue fur image classification model and the like. The training information of the tongue fur image classification model based on progressive learning is uploaded to the blockchain, so that the safety and the fair transparency to users can be ensured. The training information of the tongue fur image classification model based on progressive learning can be downloaded from the blockchain by a user so as to verify whether the training information of the tongue fur image classification model based on progressive learning is tampered or not. The blockchain referred to in this example is a novel mode of application for computer technology such as distributed data storage, point-to-point transmission, consensus mechanisms, encryption algorithms, and the like. The Blockchain (Blockchain), which is essentially a decentralised database, is a string of data blocks that are generated by cryptographic means in association, each data block containing a batch of information of network transactions for verifying the validity of the information (anti-counterfeiting) and generating the next block. The blockchain may include a blockchain underlying platform, a platform product services layer, an application services layer, and the like.
The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Among these, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technique and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend and extend human intelligence, sense the environment, acquire knowledge and use knowledge to obtain optimal results.
Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.
Example two
Referring to fig. 2, fig. 2 is a schematic structural diagram of a training device for a tongue fur image classification model based on progressive learning according to an embodiment of the present invention. As shown in fig. 2, the training device based on the progressively learning tongue fur image classification model may include:
an acquisition module 201, configured to acquire an original training data set including at least a plurality of tongue fur images;
a determining module 202, configured to determine a target resolution set according to a preset progressive learning resolution amplification manner;
the determining module 202 is further configured to determine a regularized intensity value set according to a preset progressive learning regularized intensity amplification manner, where regularized intensity values in the regularized intensity value set are in one-to-one correspondence with target resolutions in the target resolution set, and regularized intensity values corresponding to the target resolutions increase along with the increase of the target resolutions;
the resolution resetting module 203 is configured to perform a preset resolution resetting operation on each tongue fur image in the original training data set to obtain a target training data set, where the target training data set in the target training data set corresponds to the target resolution in the target resolution set one by one, and the resolution of the tongue fur image in each set of target training data set is the target resolution corresponding to the set of target training data set;
the training module 204 is configured to perform multiple training on a preset tongue fur image classification model based on the target training data set and the regularized intensity value set, so as to converge a target loss function of the tongue fur image classification model, so as to complete training on the tongue fur image classification model.
For a specific description of the training device based on the tongue fur image classification model of progressive learning, reference may be made to a specific description of the training method based on the tongue fur image classification model of progressive learning, and for avoiding repetition, details are not repeated here.
Example III
Referring to fig. 3, fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the invention. As shown in fig. 3, the computer device may include:
a memory 301 storing executable program code;
a processor 302 connected to the memory 301;
the processor 302 invokes the executable program code stored in the memory 301 to perform the steps in the training method of the tongue fur image classification model based on progressive learning disclosed in the embodiment of the present invention.
Example IV
Referring to fig. 4, an embodiment of the present invention discloses a computer storage medium 401, where the computer storage medium 401 stores computer instructions for executing the steps in the training method of the tongue fur image classification model based on progressive learning disclosed in the embodiment of the present invention when the computer instructions are called.
The apparatus embodiments described above are merely illustrative, wherein the modules illustrated as separate components may or may not be physically separate, and the components shown as modules may or may not be physical, i.e., may be located in one place, or may be distributed over a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above detailed description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course by means of hardware. Based on such understanding, the foregoing technical solutions may be embodied essentially or in part in the form of a software product that may be stored in a computer-readable storage medium including Read-Only Memory (ROM), random-access Memory (Random Access Memory, RAM), programmable Read-Only Memory (Programmable Read-Only Memory, PROM), erasable programmable Read-Only Memory (Erasable Programmable Read Only Memory, EPROM), one-time programmable Read-Only Memory (OTPROM), electrically erasable programmable Read-Only Memory (EEPROM), compact disc Read-Only Memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disc Memory, magnetic disc Memory, tape Memory, or any other medium that can be used for computer-readable carrying or storing data.
Finally, it should be noted that: the embodiment of the invention discloses a training method, a training device, computer equipment and a training medium of a tongue fur image classification model based on progressive learning, which are disclosed in the embodiment of the invention only for the preferred embodiment of the invention, and are only used for illustrating the technical scheme of the invention, but not limiting the technical scheme; although the invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art will understand that; the technical scheme recorded in the various embodiments can be modified or part of technical features in the technical scheme can be replaced equivalently; such modifications and substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (6)
1. A training method of a tongue fur image classification model based on progressive learning, the method comprising:
acquiring an original training data set at least comprising a plurality of tongue fur images;
determining a target resolution set according to a preset progressive learning resolution amplification mode;
determining a regularized intensity value set according to a preset regularized intensity amplification mode of progressive learning, wherein regularized intensity values in the regularized intensity value set are in one-to-one correspondence with target resolutions in the target resolution set, and regularized intensity values corresponding to the target resolutions are increased along with the increase of the target resolutions;
performing a preset resolution resetting operation on each tongue fur image in the original training data set to obtain a target training data set, wherein the target training data set in the target training data set corresponds to the target resolution in the target resolution set one by one, and the resolution of the tongue fur image in each group of target training data set is the target resolution corresponding to the group of target training data set;
training a preset tongue fur image classification model for a plurality of times based on the target training data set and the regularized intensity value set, so that a target loss function of the tongue fur image classification model is converged to complete training of the tongue fur image classification model;
the determining the target resolution set according to the preset progressive learning resolution amplification mode comprises the following steps:
determining the maximum resolution and the minimum resolution corresponding to the target resolution set;
determining a plurality of target resolutions from a section between the maximum resolution and the minimum resolution according to a preset target resolution determination mode, and taking the target resolutions as a target resolution set;
the determining, according to a preset target resolution determining manner, a plurality of target resolutions from a section between the maximum resolution and the minimum resolution, to be a target resolution set, includes:
determining a plurality of target resolutions from a section between the maximum resolution and the minimum resolution as a target resolution set by the following formula:
wherein S is i Is saidThe ith target resolution in the target resolution set, S 0 Is the minimum resolution, S e Is the maximum resolution, N is the total number of target resolutions in the target resolution set;
the determining the regularized intensity value set according to a preset regularized intensity amplification mode of progressive learning comprises the following steps:
determining a maximum regularization intensity value and a minimum regularization intensity value corresponding to the regularization intensity value set;
determining a plurality of regularized intensity values from intervals between the maximum regularized intensity value and the minimum regularized intensity value according to a preset regularized intensity value determining mode to serve as a regularized intensity value set;
the determining, according to a preset regularization intensity value determining manner, a plurality of regularization intensity values from a section between the maximum regularization intensity value and the minimum regularization intensity value, to be used as a regularization intensity value set, includes:
determining a plurality of regularized intensity values from the interval between the maximum regularized intensity value and the minimum regularized intensity value by the following formula to be used as a regularized intensity value set:
wherein D is i Is the ith regularized intensity value in the regularized intensity value set, which corresponds to the ith target resolution in the target resolution set, D 0 Is the minimum regularized intensity value, D e Is the maximum regularized intensity value, N is the total number of regularized intensity values in the regularized intensity value set.
2. The training method of a progressive learning based tongue fur image classification model according to claim 1, wherein the original training dataset further comprises a labeling label corresponding to each tongue fur image;
and training the preset tongue coating image classification model for a plurality of times based on the target training data set and the regularized intensity value set, so that a target loss function of the tongue coating image classification model converges, and after training the tongue coating image classification model is completed, the method further comprises:
classifying and predicting each tongue coating image in the original training data set based on the trained tongue coating image classification model to obtain a prediction label and a prediction probability value corresponding to each tongue coating image;
screening noise tongue coating images from all tongue coating images according to the prediction labels, the prediction probability values and the labeling labels corresponding to the tongue coating images, wherein the noise tongue coating images refer to tongue coating images of which the corresponding prediction labels are inconsistent with the labeling labels and the prediction probability values are lower than a preset prediction probability value threshold value;
deleting the noise tongue fur image and the labeling label corresponding to the noise tongue fur image from the original training data set so as to clean the original training data set.
3. The training method of a tongue fur image classification model based on progressive learning according to claim 1, wherein the objective loss function of the tongue fur image classification model is:
wherein y is j Labeling label corresponding to jth tongue fur image, p j The predicted probability value corresponding to the j-th tongue coating image is M, and M is the total number of the tongue coating images in the original training data set.
4. A training device for a tongue fur image classification model based on progressive learning, for implementing the training method of a tongue fur image classification model based on progressive learning as claimed in any one of claims 1-3, characterized in that the device comprises:
the acquisition module is used for acquiring an original training data set at least comprising a plurality of tongue fur images;
the determining module is used for determining a target resolution set according to a preset progressive learning resolution amplification mode;
the determining module is further configured to determine a regularized intensity value set according to a preset regularized intensity amplification manner of progressive learning, where regularized intensity values in the regularized intensity value set are in one-to-one correspondence with target resolutions in the target resolution set, and regularized intensity values corresponding to the target resolutions increase along with the increase of the target resolutions;
the resolution resetting module is used for executing preset resolution resetting operation on each tongue fur image in the original training data set to obtain a target training data set, wherein the target training data set in the target training data set corresponds to the target resolution in the target resolution set one by one, and the resolution of the tongue fur image in each group of target training data set is the target resolution corresponding to the group of target training data set;
the training module is used for training the preset tongue fur image classification model for a plurality of times based on the target training data set and the regularized intensity value set, so that the target loss function of the tongue fur image classification model is converged, and the training of the tongue fur image classification model is completed.
5. A computer device, the computer device comprising:
a memory storing executable program code;
a processor coupled to the memory;
the processor invokes the executable program code stored in the memory to perform the training method of the progressive learning based tongue fur image classification model of any of claims 1-3.
6. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements a method of training a tongue fur image classification model based on progressive learning as claimed in any one of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111086464.5A CN113780444B (en) | 2021-09-16 | 2021-09-16 | Training method of tongue fur image classification model based on progressive learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111086464.5A CN113780444B (en) | 2021-09-16 | 2021-09-16 | Training method of tongue fur image classification model based on progressive learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113780444A CN113780444A (en) | 2021-12-10 |
| CN113780444B true CN113780444B (en) | 2023-07-25 |
Family
ID=78851364
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111086464.5A Active CN113780444B (en) | 2021-09-16 | 2021-09-16 | Training method of tongue fur image classification model based on progressive learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113780444B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117094966B (en) * | 2023-08-21 | 2024-04-05 | 青岛美迪康数字工程有限公司 | Tongue image identification method and device based on image amplification and computer equipment |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109345469A (en) * | 2018-09-07 | 2019-02-15 | 苏州大学 | It is a kind of that speckle denoising method in the OCT image of confrontation network is generated based on condition |
| CN109657722A (en) * | 2018-12-20 | 2019-04-19 | 山东农业大学 | Tongue fur image-recognizing method and system based on deep learning algorithm |
| CN110738235A (en) * | 2019-09-16 | 2020-01-31 | 平安科技（深圳）有限公司 | Pulmonary tuberculosis determination method, pulmonary tuberculosis determination device, computer device, and storage medium |
| WO2021016131A1 (en) * | 2019-07-19 | 2021-01-28 | The Jackson Laboratory | Convolutional neural networks for classification of cancer histological images |
| CN112766155A (en) * | 2021-01-19 | 2021-05-07 | 山东华宇航天空间技术有限公司 | Deep learning-based mariculture area extraction method |
| CN112884648A (en) * | 2021-01-25 | 2021-06-01 | 汉斯夫(杭州)医学科技有限公司 | Method and system for multi-class blurred image super-resolution reconstruction |
| CN112991183A (en) * | 2021-04-09 | 2021-06-18 | 华南理工大学 | Video super-resolution method based on multi-frame attention mechanism progressive fusion |
| CN113379602A (en) * | 2021-06-08 | 2021-09-10 | 中国科学技术大学 | Light field super-resolution enhancement method by using zero sample learning |
| CN113392845A (en) * | 2021-06-15 | 2021-09-14 | 广州观必达数据技术有限责任公司 | Deep learning remote sensing image semantic segmentation method and system based on U-NET |
-
2021
- 2021-09-16 CN CN202111086464.5A patent/CN113780444B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109345469A (en) * | 2018-09-07 | 2019-02-15 | 苏州大学 | It is a kind of that speckle denoising method in the OCT image of confrontation network is generated based on condition |
| CN109657722A (en) * | 2018-12-20 | 2019-04-19 | 山东农业大学 | Tongue fur image-recognizing method and system based on deep learning algorithm |
| WO2021016131A1 (en) * | 2019-07-19 | 2021-01-28 | The Jackson Laboratory | Convolutional neural networks for classification of cancer histological images |
| CN110738235A (en) * | 2019-09-16 | 2020-01-31 | 平安科技（深圳）有限公司 | Pulmonary tuberculosis determination method, pulmonary tuberculosis determination device, computer device, and storage medium |
| CN112766155A (en) * | 2021-01-19 | 2021-05-07 | 山东华宇航天空间技术有限公司 | Deep learning-based mariculture area extraction method |
| CN112884648A (en) * | 2021-01-25 | 2021-06-01 | 汉斯夫(杭州)医学科技有限公司 | Method and system for multi-class blurred image super-resolution reconstruction |
| CN112991183A (en) * | 2021-04-09 | 2021-06-18 | 华南理工大学 | Video super-resolution method based on multi-frame attention mechanism progressive fusion |
| CN113379602A (en) * | 2021-06-08 | 2021-09-10 | 中国科学技术大学 | Light field super-resolution enhancement method by using zero sample learning |
| CN113392845A (en) * | 2021-06-15 | 2021-09-14 | 广州观必达数据技术有限责任公司 | Deep learning remote sensing image semantic segmentation method and system based on U-NET |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113780444A (en) | 2021-12-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111160139B (en) | Electrocardiosignal processing method and device and terminal equipment | |
| CN109919031B (en) | Human behavior recognition method based on deep neural network | |
| CN109344908B (en) | Method and apparatus for generating a model | |
| US11564612B2 (en) | Automatic recognition and classification method for electrocardiogram heartbeat based on artificial intelligence | |
| CN110335259B (en) | Medical image identification method and device and storage medium | |
| CN110443222B (en) | Method and device for training face key point detection model | |
| CN109544598A (en) | Method for tracking target, device and readable storage medium storing program for executing | |
| CN111772619A (en) | Electrocardiogram heart beat identification method, terminal device and storage medium | |
| CN112347964B (en) | Behavior detection method and device based on graph network | |
| CN112560964A (en) | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning | |
| CN110688471B (en) | Training sample obtaining method, device and equipment | |
| CN112365980A (en) | Brain tumor multi-target point auxiliary diagnosis and prospective treatment evolution visualization method and system | |
| CN110705584A (en) | Emotion recognition method, emotion recognition device, computer device and storage medium | |
| CN114998220A (en) | Tongue image detection and positioning method based on improved Tiny-YOLO v4 natural environment | |
| CN111368926A (en) | Image screening method, device and computer readable storage medium | |
| CN113780444B (en) | Training method of tongue fur image classification model based on progressive learning | |
| CN113435335A (en) | Microscopic expression recognition method and device, electronic equipment and storage medium | |
| CN114511912A (en) | Cross-library micro-expression recognition method and device based on double-current convolutional neural network | |
| CN114358249A (en) | Target recognition model training method, target recognition method and device | |
| CN111445545A (en) | Text-to-map method, device, storage medium and electronic equipment | |
| CN114386457A (en) | Electrocardiosignal abnormality detection method, system, device and storage medium | |
| CN114119151A (en) | Personalized recommendation method, system and medium for next shopping basket | |
| CN113643283A (en) | Method, device, equipment and storage medium for detecting aging condition of human body | |
| CN117033733B (en) | Intelligent automatic classification and label generation system and method for library resources | |
| CN114743043B (en) | Image classification method, electronic device, storage medium and program product |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |