CN112560964A - Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning - Google Patents
Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning Download PDFInfo
- Publication number
- CN112560964A CN112560964A CN202011506545.1A CN202011506545A CN112560964A CN 112560964 A CN112560964 A CN 112560964A CN 202011506545 A CN202011506545 A CN 202011506545A CN 112560964 A CN112560964 A CN 112560964A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- image
- chinese herbal
- data set
- 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.)
- Pending
Links
- 238000012549 training Methods 0.000 title claims abstract description 181
- 241000411851 herbal medicine Species 0.000 title claims abstract description 130
- 241000607479 Yersinia pestis Species 0.000 title claims abstract description 120
- 201000010099 disease Diseases 0.000 title claims abstract description 64
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 61
- 230000001575 pathological effect Effects 0.000 claims abstract description 88
- 238000013136 deep learning model Methods 0.000 claims abstract description 86
- 238000002372 labelling Methods 0.000 claims abstract description 74
- 230000007170 pathology Effects 0.000 claims abstract description 74
- 238000012360 testing method Methods 0.000 claims description 47
- 238000012795 verification Methods 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 16
- 238000003709 image segmentation Methods 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 7
- 241000238631 Hexapoda Species 0.000 claims description 5
- 238000010276 construction Methods 0.000 abstract 1
- 230000006870 function Effects 0.000 description 13
- 230000001580 bacterial effect Effects 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000002538 fungal effect Effects 0.000 description 4
- 239000002689 soil Substances 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 241000237852 Mollusca Species 0.000 description 2
- 208000003141 Plant Poisoning Diseases 0.000 description 2
- 241000700605 Viruses Species 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000001717 pathogenic effect Effects 0.000 description 2
- 239000003123 plant toxin Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000007670 refining Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 230000003612 virological effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Investigating Or Analysing Biological Materials (AREA)
Abstract
The invention discloses a method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning, which comprises the following steps of: acquiring an annotated image dataset comprising a plurality of sample images, each sample image being annotated with a corresponding sample pathology type label; training a Resnet50 deep learning model based on the sample image to obtain a first labeling model; acquiring an unmarked image data set comprising a plurality of unmarked images, and inputting the unmarked images into a first marked model to obtain the pathological types corresponding to the unmarked images and the probability values corresponding to the pathological types; training a finetune model corresponding to the first labeling model based on the labeled image data set, the pathological types of the unlabeled images and the probability values corresponding to the labeled image data set and the unlabeled images to obtain a Chinese herbal medicine pest and disease identification model; the Chinese herbal medicine image to be marked is identified through the Chinese herbal medicine pest identification model, and the identification accuracy is improved. The invention relates to an intelligent medical scene, and therefore the construction of an intelligent city is promoted.
Description
Technical Field
The embodiment of the invention relates to the technical field of artificial intelligence, in particular to a method and a system for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning.
Background
At present, deep learning is an algorithm sharer in the big data era and becomes a research hotspot in recent years. Compared with the traditional artificial intelligence algorithm, the deep learning technology has two advantages. One is that the deep learning technique can continuously improve the performance of the data as the scale of the data increases, and the traditional artificial intelligence algorithm (including a rule-based expert system) is difficult to continuously improve the performance of the data by using mass data. Secondly, the deep learning technology can directly extract features from data, so that the work of designing a feature extractor for each problem is reduced, and the traditional artificial intelligence algorithm needs to manually extract the features, for example, a traditional medical expert system needs to extract expert rules based on the data. The advantages of deep learning techniques have been well developed in some fields, for example, image classification techniques based on deep convolutional networks have exceeded the accuracy of the human eye, speech recognition techniques based on deep neural networks have reached 95% accuracy, and machine translation techniques based on deep neural networks have approached the average translation level of humans.
In the aspect of Chinese herbal medicine pest image data annotation, the Chinese herbal medicine pest image data annotation work is time-consuming, labor cost input is large, and cost is high; the Chinese herbal medicine pest image data annotation needs Chinese herbal medicine experts with rich experience, and the annotation difficulty is high; and a small amount of labeled data cannot meet the identification precision requirement.
Due to the problems, the deep learning technology is lack of a mature and effective system in the aspect of Chinese herbal medicine pest image data annotation at present, and multisource irregular Chinese herbal medicine pest image data annotation is not effectively utilized.
Disclosure of Invention
In view of this, an object of the embodiments of the present invention is to provide a method and a system for training a Chinese herbal medicine pest recognition model based on semi-supervised learning, so as to improve the training accuracy of the Chinese herbal medicine pest recognition model and obtain a more accurate recognition result when in application.
In order to achieve the purpose, the embodiment of the invention provides a method for training a Chinese herbal medicine disease and insect pest recognition model based on semi-supervised learning, which comprises the following steps:
acquiring an annotated image dataset which comprises a plurality of sample images, wherein each sample image has a sample pathology type label corresponding to the sample image;
training a Resnet50 deep learning model based on the sample images in the labeled image dataset to obtain a first labeled model;
acquiring an unlabelled image data set, and inputting a plurality of unlabelled images contained in the unlabelled image set into the first labeling model to obtain a pathological type corresponding to each unlabelled image and a probability value corresponding to the pathological type;
combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set, and training a finetune model corresponding to the first labeled model based on the new training data set to obtain a Chinese herbal medicine pest and disease identification model;
acquiring a Chinese herbal medicine image to be labeled, and inputting the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest and disease identification model to obtain a target pathological type of the Chinese herbal medicine image and a target probability value corresponding to the target pathological type.
Further, combining the labeled image dataset with the pathology types corresponding to the unlabeled images and the probability values corresponding to the pathology types into a new training dataset, training a finetune model corresponding to the first labeled model based on the new training dataset to obtain a Chinese herbal medicine pest and disease identification model includes:
combining the labeled image data set, the pathology types corresponding to the unlabeled images and the probability values corresponding to the pathology types into a new training data set;
training a finetune model corresponding to the first labeling model based on the new training data set, wherein in the process of training the finetune model, the loss weight value of the labeling image data set is set as a first weight value, and the loss weight value of the unlabeled image is set as a second weight value according to the probability value, so that the Chinese herbal medicine pest and disease identification model is obtained.
Further, training a Resnet50 deep learning model based on the sample images in the annotation image dataset to obtain a first annotation model includes:
dividing the annotated image dataset into a training dataset, a verification dataset and a test dataset;
taking the sample images in the training data set as the input of the Resnet50 deep learning model, taking the probability values of the sample images in the training data set and the corresponding pathological types as the output of the Resnet50 deep learning model, and training the Resnet50 deep learning model to obtain a second labeling model;
and inputting the sample images in the verification data set into the second labeling model for verification, and correcting the second labeling model according to a verification result to obtain a first labeling model.
Further, the inputting of the sample images in the training data set as the Resnet50 deep learning model, and the outputting of the probability values of the sample images in the training data set and the corresponding pathology types as the Resnet50 deep learning model, wherein the training of the Resnet50 deep learning model includes:
inputting sample images in the training dataset into the Resnet50 deep learning model;
carrying out image segmentation processing on the sample image through the Resnet50 deep learning model to obtain a plurality of mutually disjoint to-be-detected region images;
extracting image contour features of each to-be-detected region image, and calculating a probability value between the sample image and a pathological type according to the extracted image contour features;
and taking the maximum probability value in the pathology types as an output pathology type, and adjusting the Resnet50 deep learning model according to the difference value between the output pathology type and the sample pathology type corresponding to the sample image.
Further, the training of the Resnet50 deep learning model with the sample images in the training data set as the input of the Resnet50 deep learning model and the probability values of the sample images in the training data set and the corresponding pathology types as the output of the Resnet50 deep learning model to obtain a second labeling model includes:
taking the sample images in the training data set as the Resnet50 deep learning model input, training the Resnet50 deep learning model to output probability values of the sample images and the pathological types corresponding to the sample images;
and adjusting parameters of the Resnet50 deep learning model according to the difference value of the output probability value and the probability value corresponding to the sample image in the training data set, continuing to train the Resnet50 deep learning model until the Resnet50 deep learning model converges, and taking the Resnet50 deep learning model at the time of convergence as the second labeling model.
Further, the training of the Resnet50 deep learning model with the sample images in the training data set as the input of the Resnet50 deep learning model and the probability values of the sample images in the training data set and the corresponding pathology types as the output of the Resnet50 deep learning model to obtain a second labeling model includes:
inputting the test data set into the trained second labeling model to judge whether the second labeling model outputs the test pathology type of the test data set and the test probability value corresponding to the test medical record type;
and if the test pathology type of the test data set and the test probability value corresponding to the test medical record type are output, the second labeling model is successfully trained.
Further, the method further comprises:
and uploading the Chinese herbal medicine pest and disease damage identification model to a block chain.
In order to achieve the above object, an embodiment of the present invention provides a system for training a Chinese herbal medicine pest recognition model based on semi-supervised learning, including:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring an annotated image dataset which comprises a plurality of sample images, and each sample image is provided with a sample pathology type label corresponding to the sample image;
the first training module is used for training a Resnet50 deep learning model based on the sample images in the annotation image dataset to obtain a first annotation model;
the second acquisition module is used for acquiring an unlabeled image data set, inputting a plurality of unlabeled images contained in the unlabeled image set into the first labeling model, and obtaining a pathological type corresponding to each unlabeled image and a probability value corresponding to the pathological type;
the second training module is used for combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set, and training the finetune model corresponding to the first labeled model based on the new training data set to obtain a Chinese herbal medicine pest and disease identification model;
the identification module is used for acquiring a Chinese herbal medicine image to be labeled and inputting the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest and disease identification model so as to obtain a target pathological type of the Chinese herbal medicine image and a target probability value corresponding to the target pathological type.
In order to achieve the above object, an embodiment of the present invention provides a computer device, which includes a memory and a processor, where the memory stores a computer program that can be executed on the processor, and the computer program, when executed by the processor, implements the steps of the method for training the Chinese herbal medicine pest and disease identification model based on semi-supervised learning as described above.
To achieve the above object, an embodiment of the present invention provides a computer-readable storage medium, having a computer program stored therein, where the computer program is executable by at least one processor to cause the at least one processor to execute the steps of the method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning as described above.
The method and the system for training the Chinese herbal medicine pest and disease identification model based on semi-supervised learning provided by the embodiment of the invention have the advantages that the Resnet50 deep learning model is trained through the labeled image data set to obtain a first labeled model, training is carried out on training data, test data and verification data during training, further testing is carried out through an unlabelled image data set, the test result is used for readjusting the first labeled model to obtain a second labeled model, finally, the labeled image data set and the unlabelled image data set are combined into a new training data set, and the second labeled model is retrained to obtain the Chinese herbal medicine pest and disease identification model, so that the training accuracy of the Chinese herbal medicine pest and disease identification model is improved, and a more accurate identification result is obtained during application.
Drawings
FIG. 1 is a flow chart of a first embodiment of a method for training a Chinese herbal medicine pest identification model based on semi-supervised learning.
FIG. 2 is a schematic diagram of program modules of a second embodiment of the system for training a Chinese herbal medicine pest identification model based on semi-supervised learning.
Fig. 3 is a schematic diagram of a hardware structure of a third embodiment of the computer device according to the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example one
Referring to fig. 1, a flowchart of steps of a method for training a Chinese herbal medicine pest recognition model based on semi-supervised learning according to a first embodiment of the present invention is shown. It is to be understood that the flow charts in the embodiments of the present method are not intended to limit the order in which the steps are performed. The following description is made by way of example with the computer device 2 as the execution subject. The details are as follows.
Step S100, an annotated image dataset is obtained, wherein the annotated image dataset comprises a plurality of sample images, and each sample image has a sample pathology type label corresponding to the sample image.
Specifically, the marked image dataset is marked Chinese herbal medicine pest image data, the sample image is a Chinese herbal medicine pest image, the sample pathological type is a pest type and comprises a disease type and a physiological type, the physiological type comprises a natural disaster type, a deficiency type, a climate factor type, a plant poisoning type and a pest type, the disease type comprises a bacterial type, a fungal type and a virus bacterial type, and the pest type comprises a soil pest type, a ground pest type, a mite pest type and a pest soft body type. The pathological type of the sample can be specifically disease type and pest type. And when the image is labeled, manually labeling, firstly, preprocessing the sample image such as denoising, and labeling the preprocessed sample image to the corresponding disease and insect pest type.
Step S120, training a Resnet50 deep learning model based on the sample images in the annotation image dataset to obtain a first annotation model.
Specifically, the Resnet50 deep learning model is a deep learning model with a residual error network of 50, and a deep learning model weight file of the residual error network Resnet50 can be used as a pre-training model to improve learning efficiency.
Exemplarily, the step S120 further includes:
step S121, dividing the annotation image data set into a training data set, a verification data set and a test data set.
Specifically, the marked Chinese herbal medicine pest image data is calculated according to the following steps of 8: 1: the method comprises the following steps of 1, dividing the Chinese herbal medicine image data into a training data set, a verification data set and a test data set, verifying a labeling model if the verification data set is Chinese herbal medicine image data which is not labeled, outputting a pathology type corresponding to a maximum probability value based on the labeling model, and manually verifying a result. The test data set is the Chinese herbal medicine image data with part marked, and the test process is similar to the verification process.
Step S122, using the sample images in the training data set as the input of the Resnet50 deep learning model, and using the probability values of the sample images in the training data set and the corresponding pathology types as the output of the Resnet50 deep learning model, training the Resnet50 deep learning model, and obtaining a second labeled model.
Specifically, a Resnet50 deep learning model is trained according to a training data set, and the sample images are imported into the deep learning model for prediction operation to obtain corresponding pathological types and probability values of the pathological types. Calculating probability values of the sample images through a softmax function deeply learned by Resnet 50; and selecting the sample pathology type corresponding to the maximum probability value, determining the sample pathology type corresponding to the sample image, and outputting the corresponding sample probability value. And training the Resnet50 deep learning model so that the Resnet50 deep learning model can output the corresponding sample pathology type and the probability value of the sample type according to the input sample image to obtain a second labeling model.
And S123, inputting the sample image in the verification data set into the annotation model for verification, and correcting the second annotation model according to a verification result to obtain a first annotation model.
Specifically, the verification data set is input into the labeling model, the output sample images labeled with the pathological types and the sample images not labeled with the pathological types are checked for correctness, and the loss weight value of the second labeling model is modified until the probability value of the pathological types corresponding to the output test data reaches more than 98%, and the correctness of the output pathological types also reaches 98%.
Exemplarily, the step S122 further includes:
step S122A, inputting the sample images in the training dataset into the Resnet50 deep learning model.
Specifically, a single sample image is input to the Resnet50 deep learning model and identified by the Resnet50 deep learning model.
And step S122B, carrying out image segmentation processing on the sample image through the Resnet50 deep learning model to obtain a plurality of mutually disjoint images of the region to be detected.
Specifically, when the labeling model training is performed based on the training data, the image segmentation processing is performed on the sample image to obtain a plurality of mutually disjoint to-be-detected region images. Since the sample image generally contains one or more images of Chinese herbal medicines, the image of a single Chinese herbal medicine can be limited in an image of a region to be detected (part of the image of the region to be detected may not contain the image of the Chinese herbal medicine) through image segmentation processing, so that the subsequent effective identification of the disease and insect pest type of the single Chinese herbal medicine can be carried out. The image segmentation processing method may be, but not limited to, a threshold-based image segmentation method, a region growing-based image segmentation method, a deformation model-based image segmentation method, a graph theory-based image segmentation method, a clustering-based image segmentation method, or a classification-based image segmentation method, and the like.
Step S122C, extracting image contour features of each to-be-detected region image, and calculating a probability value between the sample image and a pathology type according to the extracted image contour features.
For each image of the area to be detected, firstly extracting image contour features, and then if corresponding Chinese herbal medicines are identified according to the extracted image contour features, the Chinese herbal medicines can be represented by sequentially connecting coordinates of a plurality of key feature points in series to form a shape vector, so that after the image of the area to be detected is processed, the corresponding image contour features can be extracted. In order to identify the corresponding Chinese herbal medicine according to the extracted image contour features, standard image contour features of a plurality of different Chinese herbal medicines are prepared for comparison and identification, namely the method further comprises the following steps: the method comprises the steps of obtaining standard images of various Chinese herbal medicines, and then extracting corresponding standard image contour features aiming at the standard images of each Chinese herbal medicine.
Step S122D, using the most probable value in the pathology types as an output pathology type, and adjusting the Resnet50 deep learning model according to a difference value between the output pathology type and a sample pathology type corresponding to the sample image.
Specifically, calculating a probability value between the image of the region to be detected and the sample pathology type through a softmax function deeply learned by Resnet 50; and selecting the sample pathology type corresponding to the maximum probability value, determining the sample pathology type corresponding to the sample image as the output pathology type of the corresponding sample image, and outputting the corresponding probability value. The softmax function can be adjusted through the probability value, and the weight value of loss characteristic loss deeply learned by Resnet50 can be adjusted through outputting the difference value between the pathology type and the sample pathology type.
Exemplarily, the step S122 further includes:
taking the sample images in the training data set as the Resnet50 deep learning model input, training the Resnet50 deep learning model to output probability values of the sample images and the pathological types corresponding to the sample images;
and adjusting parameters of the Resnet50 deep learning model according to the difference value of the output probability value and the probability value corresponding to the sample image in the training data set, continuing to train the Resnet50 deep learning model until the Resnet50 deep learning model converges, and taking the Resnet50 deep learning model at the time of convergence as the second labeling model.
Specifically, the output probability value of the trained Resnet50 deep learning model is close to the probability value corresponding to the sample image, the probability value corresponding to the sample image is set to 1, but the accuracy cannot be achieved during training, the probability value can be set to be more than 95%, the output probability value reaches more than 95%, and the sample pathology type corresponding to the output probability value is correct, so that the training convergence of the Resnet50 deep learning model is represented, and the Resnet50 deep learning model during convergence is used as the second labeling model.
After step S122, the method further includes:
inputting the test data set into the trained second labeling model to judge whether the second labeling model outputs the test pathology type of the test data set and the test probability value corresponding to the test medical record type;
and if the test pathology type of the test data set and the test probability value corresponding to the test medical record type are output, the second labeling model is successfully trained.
Specifically, test data is input into a second labeling model obtained after training, whether an output result is a test pathology type and a test probability value corresponding to a test case history type is judged, whether the output result reaches a result accuracy rate that a correct output result accounts for a total output result is judged, if not, the step of training the second labeling model through a training data set is repeated until the result accuracy rate is reached, and the second labeling model is obtained.
Step S140, obtaining an unmarked image data set, inputting a plurality of unmarked images contained in the unmarked image data set into the first annotated model, and obtaining a pathological type corresponding to each unmarked image and a probability value corresponding to the pathological type.
Specifically, a large number of unlabelled image data sets are obtained, and the unlabelled image data sets are identified through a first labeling model, so that the pathological type of the unlabelled image and the probability value corresponding to the pathological type are obtained. The first labeling model can output the pathology type with the maximum probability value through the input unlabeled image as the pathology type of the unlabeled image, and simultaneously output the probability value.
And selecting the sample pathology type corresponding to the maximum probability value, determining the sample pathology type as the pathology type of the unlabeled image, and outputting the corresponding probability value.
Illustratively, the softmax function is as follows: softmax (x)i＝exp(xi)/∑jexp(xj) Wherein x represents the sample pathology type.
The output classification vector is (0.20, 0.48, 0.66, 0.95, 0.80, …), which means:
when the pathological type of the sample is A, the probability is 0.20;
when the pathological type of the sample is B, the probability is 0.48;
when the pathological type of the sample is C, the probability is 0.66;
when the pathological type of the sample is D, the probability is 0.95;
when the pathological type of the sample is E, the probability is 0.80;
and so on, without exhaustive enumeration.
The computer device selects the probability value with the highest probability to determine the probability value of the unlabeled image, namely the probability value of the unlabeled image is 4, and the corresponding sample pathology type is D. A. B, C, D, E, etc. are sample pathological types including pathogenic types of bacterial type, fungal type and viral type; is a disease and pest type of soil pest type, ground pest type, mite pest type and mollusk pest type.
And S160, combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set, and training the finetune model corresponding to the first labeled model based on the new training data set to obtain the Chinese herbal medicine pest and disease identification model.
Specifically, the output pathological type of the unmarked image may not be accurate enough, the pathological type of the marked image data set is standard, the loss weight values of the unmarked image data set and the marked image data set are modified to fine-tune the loss weight of the first marked model, and then the Chinese herbal medicine pest and disease identification model is obtained through training to improve the output accuracy of the Chinese herbal medicine pest and disease identification model.
Exemplarily, the step S160 further includes:
and combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set.
Specifically, a first labeling model is adjusted and trained, wherein a first weight value of a labeled image data set in the loss is modified to be 1.0, and a second weight value of an unlabeled image data set in the loss is a probability value, so that the first labeling model is trained to obtain the Chinese herbal medicine pest and disease identification model. The finetune model includes: the new data (unlabeled image dataset) and the base data (labeled image data) are to have a correlation; the parameters of the first labeling model can be fixed firstly, and the learning rate of a newly added layer needs to be properly increased; parameters of the first annotation model are gradually released for training, but the accuracy rate is controlled. The first labeling model is repeatedly trained until the optimal solution of the model is obtained, which can be understood that the probability value of the output of the model is more than 95%. The main training steps for training the first label model based on the new training data set are as follows: improving the accuracy of the Chinese herbal medicine pest and disease identification model based on the unlabeled data; the weighted value of the loss of the unlabeled data is according to the type probability value output by the basic model; and (5) iteratively and repeatedly training the model in a finetune mode.
Training a finetune model corresponding to the first labeling model based on the new training data set, wherein in the process of training the finetune model, the loss weight value of the labeling image data set is set as a first weight value, and the loss weight value of the unlabeled image is set as a second weight value according to the probability value, so that the Chinese herbal medicine pest and disease identification model is obtained.
Specifically, the weight value of the loss function of the loss in the training process of the first labeling model is adjusted, the first weight value of the labeled image data set in the loss is modified to 1.0, and the second weight value of the unlabeled image data set in the loss is the probability value output in the step. Merging the marked image data and the unmarked image data set into a new training data set, then refining the first marked model, namely finely adjusting the first marked model, reserving the model data of the first marked model, and adjusting the weight value of loss in the model training process to obtain the Chinese herbal medicine pest and disease damage identification model.
Step S180, obtaining a Chinese herbal medicine image to be labeled, and inputting the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest and disease identification model to obtain a target pathological type of the Chinese herbal medicine image and a target probability value corresponding to the target pathological type.
Specifically, the Chinese herbal medicine image to be labeled is input into a Chinese herbal medicine pest and disease identification model, so that the target pathological type of the Chinese herbal medicine image and the target probability value corresponding to the target pathological type are output through the Chinese herbal medicine pest and disease identification model.
Illustratively, the method further comprises:
and uploading the Chinese herbal medicine pest and disease damage identification model to a block chain.
Specifically, the Chinese herbal medicine pest and disease identification model is uploaded to the block chain, so that the safety and the fair transparency of the Chinese herbal medicine pest and disease identification model to users can be guaranteed. The user equipment can download the Chinese herbal medicine pest identification model from the block chain so as to check whether the Chinese herbal medicine pest identification model is tampered. The blockchain referred to in this example is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, consensus mechanism, encryption algorithm, and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.
Example two
Referring to fig. 2, a schematic diagram of a program module of a second embodiment of the system for training a Chinese herbal medicine pest identification model based on semi-supervised learning according to the present invention is shown. In this embodiment, the system 20 for training a Chinese herbal medicine pest recognition model based on semi-supervised learning may include or be divided into one or more program modules, and the one or more program modules are stored in a storage medium and executed by one or more processors to complete the present invention and implement the above-mentioned method for training a Chinese herbal medicine pest recognition model based on semi-supervised learning. The program module referred to in the embodiment of the present invention is a series of computer program instruction segments capable of performing specific functions, and is more suitable for describing the execution process of the system 20 for training the Chinese herbal medicine pest identification model based on semi-supervised learning in a storage medium than the program itself. The following description will specifically describe the functions of the program modules of the present embodiment:
a first obtaining module 200, configured to obtain an annotated image dataset, where the annotated image dataset includes a plurality of sample images, and each sample image has a sample pathology type label corresponding to the sample image.
Specifically, the marked image dataset is marked Chinese herbal medicine pest image data, the sample image is a Chinese herbal medicine pest image, the sample pathological type is a pest type and comprises a disease type and a physiological type, the physiological type comprises a natural disaster type, a deficiency type, a climate factor type, a plant poisoning type and a pest type, the disease type comprises a bacterial type, a fungal type and a virus bacterial type, and the pest type comprises a soil pest type, a ground pest type, a mite pest type and a pest soft body type. The pathological type of the sample can be specifically disease type and pest type. And when the image is labeled, manually labeling, firstly, preprocessing the sample image such as denoising, and labeling the preprocessed sample image to the corresponding disease and insect pest type.
A first training module 202, configured to train a Resnet50 deep learning model based on the sample images in the annotation image dataset, to obtain a first annotation model.
Specifically, the Resnet50 deep learning model is a deep learning model with a residual error network of 50, and a deep learning model weight file of the residual error network Resnet50 can be used as a pre-training model to improve learning efficiency.
Illustratively, the first training module 202 is specifically configured to:
the annotated image dataset is divided into a training dataset, a validation dataset, and a test dataset.
Specifically, the marked Chinese herbal medicine pest image data is calculated according to the following steps of 8: 1: the method comprises the following steps of 1, dividing the Chinese herbal medicine image data into a training data set, a verification data set and a test data set, verifying a labeling model if the verification data set is Chinese herbal medicine image data which is not labeled, outputting a pathology type corresponding to a maximum probability value based on the labeling model, and manually verifying a result. The test data set is the Chinese herbal medicine image data with part marked, and the test process is similar to the verification process.
And taking the sample images in the training data set as the input of the Resnet50 deep learning model, taking the probability values of the sample images in the training data set and the corresponding pathological types as the output of the Resnet50 deep learning model, and training the Resnet50 deep learning model to obtain a second labeling model.
Specifically, a Resnet50 deep learning model is trained according to a training data set, and the sample images are imported into the deep learning model for prediction operation to obtain corresponding pathological types and probability values of the pathological types. Calculating probability values of the sample image through a softmax function deeply learned by Resnet 50; and selecting the sample pathology type corresponding to the maximum probability value, determining the sample pathology type corresponding to the sample image, and outputting the corresponding sample probability value. And training the Resnet50 deep learning model so that the Resnet50 deep learning model can output the corresponding sample pathology type and the probability value of the sample type according to the input sample image to obtain a second labeling model. Verification by verification data
And inputting the sample image in the verification data set into the annotation model for verification, and correcting the second annotation model according to a verification result to obtain a first annotation model.
Specifically, the verification data set is input into the labeling model, the output sample images labeled with the pathological types and the sample images not labeled with the pathological types are checked for correctness, and the loss weight value of the second labeling model is modified until the probability value of the pathological types corresponding to the output test data reaches more than 98%, and the correctness of the output pathological types also reaches 98%. And inputting the test data into the second labeling model after the first correction, and judging whether the output result reaches the result accuracy rate that the correct output result accounts for the total output result, if not, repeating the steps of training and modifying the second labeling model through the training data set and the verification data set until the result accuracy rate is reached to obtain the first labeling model.
The second obtaining module 204 is configured to obtain an unlabeled image dataset, and input a plurality of unlabeled images included in the unlabeled image dataset into the first labeling model to obtain a pathology type corresponding to each unlabeled image and a probability value corresponding to the pathology type.
Specifically, a large number of unlabelled image data sets are obtained, and the unlabelled image data sets are identified through a first labeling model, so that the pathological type of the unlabelled image and the probability value corresponding to the pathological type are obtained. The first labeling model can output the pathology type with the maximum probability value through the input unlabeled image as the pathology type of the unlabeled image, and simultaneously output the probability value.
And selecting the sample pathology type corresponding to the maximum probability value, determining the sample pathology type as the pathology type of the unlabeled image, and outputting the corresponding probability value.
Illustratively, the softmax function is as follows: softmax (x)i＝exp(xi)/∑jexp(xj) Wherein x represents the sample pathology type.
The output classification vector is (0.20, 0.48, 0.66, 0.95, 0.80, …), which means:
when the pathological type of the sample is A, the probability is 0.20;
when the pathological type of the sample is B, the probability is 0.48;
when the pathological type of the sample is C, the probability is 0.66;
when the pathological type of the sample is D, the probability is 0.95;
when the pathological type of the sample is E, the probability is 0.80;
and so on, without exhaustive enumeration.
The computer device selects the probability value with the highest probability to determine the probability value of the unlabeled image, namely the probability value of the unlabeled image is 4, and the corresponding sample pathology type is D. A. B, C, D, E, etc. are sample pathological types including pathogenic types of bacterial type, fungal type and viral type; is a disease and pest type of soil pest type, ground pest type, mite pest type and mollusk pest type.
And the second training module 206 is configured to combine the labeled image dataset and the pathology types corresponding to the unlabeled images and the probability values corresponding to the pathology types into a new training dataset, and train the finetune model corresponding to the first labeled model based on the new training dataset to obtain a Chinese herbal medicine pest and disease identification model.
Specifically, the output pathological type of the unmarked image may not be accurate enough, the pathological type of the marked image data set is standard, the loss weight values of the unmarked image data set and the marked image data set are modified to fine-tune the loss weight of the first marked model, and then the Chinese herbal medicine pest and disease identification model is obtained through training to improve the output accuracy of the Chinese herbal medicine pest and disease identification model.
Illustratively, the second training module 206 is specifically configured to:
and combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set.
Specifically, a first labeling model is adjusted and trained, wherein a first weight value of a labeled image data set in the loss is modified to be 1.0, and a second weight value of an unlabeled image data set in the loss is a probability value, so that the first labeling model is trained to obtain the Chinese herbal medicine pest and disease identification model. The finetune model includes: the new data (unlabeled image dataset) and the base data (labeled image data) are to have a correlation; the parameters of the first labeling model can be fixed firstly, and the learning rate of a newly added layer needs to be properly increased; parameters of the first annotation model are gradually released for training, but the accuracy rate is controlled. The first labeling model is repeatedly trained until the optimal solution of the model is obtained, which can be understood that the probability value of the output of the model is more than 95%. The main training steps for training the first label model based on the new training data set are as follows: improving the accuracy of the Chinese herbal medicine pest and disease identification model based on the unlabeled data; the weighted value of the loss of the unlabeled data is according to the type probability value output by the basic model; and (5) iteratively and repeatedly training the model in a finetune mode.
Training a finetune model corresponding to the first labeling model based on the new training data set, wherein in the process of training the finetune model, the loss weight value of the labeling image data set is set as a first weight value, and the loss weight value of the unlabeled image is set as a second weight value according to the probability value, so that the Chinese herbal medicine pest and disease identification model is obtained.
Specifically, the weight value of the loss function of the loss in the training process of the first labeling model is adjusted, the first weight value of the labeled image data set in the loss is modified to 1.0, and the second weight value of the unlabeled image data set in the loss is the probability value output in the step. Merging the marked image data and the unmarked image data set into a new training data set, then refining the first marked model, namely finely adjusting the first marked model, reserving the model data of the first marked model, and adjusting the weight value of loss in the model training process to obtain the Chinese herbal medicine pest and disease damage identification model.
The identification module 208 is configured to acquire a Chinese herbal medicine image to be labeled, and input the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest identification model to obtain a target pathology type of the Chinese herbal medicine image and a target probability value corresponding to the target pathology type.
Specifically, the Chinese herbal medicine image to be labeled is input into a Chinese herbal medicine pest and disease identification model, so that the target pathological type of the Chinese herbal medicine image and the target probability value corresponding to the target pathological type are output through the Chinese herbal medicine pest and disease identification model.
EXAMPLE III
Fig. 3 is a schematic diagram of a hardware architecture of a computer device according to a third embodiment of the present invention. In the present embodiment, the computer device 2 is a device capable of automatically performing numerical calculation and/or information processing in accordance with a preset or stored instruction. The computer device 2 may be a rack server, a blade server, a tower server or a rack server (including an independent server or a server cluster composed of a plurality of servers), and the like. As shown in fig. 3, the computer device 2 at least includes, but is not limited to, a memory 21, a processor 22, a network interface 23, and a system 20 for training a Chinese herbal medicine pest recognition model based on semi-supervised learning, which are connected in communication with each other through a system bus. Wherein:
in this embodiment, the memory 21 includes at least one type of computer-readable storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. In some embodiments, the storage 21 may be an internal storage unit of the computer device 2, such as a hard disk or a memory of the computer device 2. In other embodiments, the memory 21 may also be an external storage device of the computer device 2, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), or the like provided on the computer device 2. Of course, the memory 21 may also comprise both internal and external memory units of the computer device 2. In this embodiment, the memory 21 is generally used to store an operating system and various application software installed on the computer device 2, such as the program code of the system 20 for training the Chinese herbal medicine pest identification model based on semi-supervised learning in the second embodiment. Further, the memory 21 may also be used to temporarily store various types of data that have been output or are to be output.
The network interface 23 may comprise a wireless network interface or a wired network interface, and the network interface 23 is generally used for establishing communication connection between the server 2 and other electronic devices. For example, the network interface 23 is used to connect the server 2 to an external terminal via a network, establish a data transmission channel and a communication connection between the server 2 and the external terminal, and the like. The network may be a wireless or wired network such as an Intranet (Intranet), the Internet (Internet), a Global System of Mobile communication (GSM), Wideband Code Division Multiple Access (WCDMA), a 4G network, a 5G network, Bluetooth (Bluetooth), Wi-Fi, and the like. It is noted that fig. 3 only shows the computer device 2 with components 20-23, but it is to be understood that not all shown components are required to be implemented, and that more or less components may be implemented instead.
In this embodiment, the system 20 for training chinese herbal medicine pest recognition model based on semi-supervised learning stored in the memory 21 can be further divided into one or more program modules, and the one or more program modules are stored in the memory 21 and executed by one or more processors (in this embodiment, the processor 22) to complete the present invention.
For example, fig. 2 is a schematic diagram of program modules of a second embodiment of the system 20 for training a Chinese herbal medicine pest identification model based on semi-supervised learning, in which the system 20 for training a Chinese herbal medicine pest identification model based on semi-supervised learning can be divided into a first obtaining module 200, a first training module 202, a second obtaining module 204, a second training module 206 and an identification module 208. The program module referred to in the present invention refers to a series of computer program instruction segments capable of performing specific functions, and is more suitable for describing the execution process of the system 20 for training the Chinese herbal medicine pest identification model based on semi-supervised learning in the computer device 2 than a program. The specific functions of the program modules 200 and 208 have been described in detail in the second embodiment, and are not described herein again.
Example four
The present embodiment also provides a computer-readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application mall, etc., on which a computer program is stored, which when executed by a processor implements corresponding functions. The computer-readable storage medium of this embodiment is used for a computer program, and when executed by a processor, the method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning of the first embodiment is implemented.
The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.
The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning is characterized by comprising the following steps:
acquiring an annotated image dataset which comprises a plurality of sample images, wherein each sample image has a sample pathology type label corresponding to the sample image;
training a Resnet50 deep learning model based on the sample images in the labeled image dataset to obtain a first labeled model;
acquiring an unlabelled image data set, and inputting a plurality of unlabelled images contained in the unlabelled image set into the first labeling model to obtain a pathological type corresponding to each unlabelled image and a probability value corresponding to the pathological type;
combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set, and training a finetune model corresponding to the first labeled model based on the new training data set to obtain a Chinese herbal medicine pest and disease identification model;
acquiring a Chinese herbal medicine image to be labeled, and inputting the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest and disease identification model to obtain a target pathological type of the Chinese herbal medicine image and a target probability value corresponding to the target pathological type.
2. The method for training a Chinese herbal medicine pest recognition model based on semi-supervised learning according to claim 1, wherein the combining the labeled image dataset with the pathology types corresponding to the unlabeled images and the probability values corresponding to the pathology types into a new training dataset, and training the finetune model corresponding to the first labeled model based on the new training dataset to obtain the Chinese herbal medicine pest recognition model comprises:
combining the labeled image data set, the pathology types corresponding to the unlabeled images and the probability values corresponding to the pathology types into a new training data set;
training a finetune model corresponding to the first labeling model based on the new training data set, wherein in the process of training the finetune model, the loss weight value of the labeling image data set is set as a first weight value, and the loss weight value of the unlabeled image is set as a second weight value according to the probability value, so that the Chinese herbal medicine pest and disease identification model is obtained.
3. The method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning of claim 1, wherein the training of the Resnet50 deep learning model based on the sample images in the labeled image dataset to obtain a first labeled model comprises:
dividing the annotated image dataset into a training dataset, a verification dataset and a test dataset;
taking the sample images in the training data set as the input of the Resnet50 deep learning model, taking the probability values of the sample images in the training data set and the corresponding pathological types as the output of the Resnet50 deep learning model, and training the Resnet50 deep learning model to obtain a second labeling model;
and inputting the sample images in the verification data set into the second labeling model for verification, and correcting the second labeling model according to a verification result to obtain a first labeling model.
4. The method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning of claim 3, wherein the sample images in the training data set are input as the Resnet50 deep learning model, the probability values of the sample images in the training data set and the corresponding pathological types are output as the Resnet50 deep learning model, and the training of the Resnet50 deep learning model comprises the following steps:
inputting sample images in the training dataset into the Resnet50 deep learning model;
carrying out image segmentation processing on the sample image through the Resnet50 deep learning model to obtain a plurality of mutually disjoint to-be-detected region images;
extracting image contour features of each to-be-detected region image, and calculating a probability value between the sample image and a pathological type according to the extracted image contour features;
and taking the maximum probability value in the pathology types as an output pathology type, and adjusting the Resnet50 deep learning model according to the difference value between the output pathology type and the sample pathology type corresponding to the sample image.
5. The method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning of claim 4, wherein the step of training the Resnet50 deep learning model by using the sample images in the training data set as the Resnet50 deep learning model input and the probability values of the sample images in the training data set and the corresponding pathological types as the Resnet50 deep learning model output comprises the steps of:
taking the sample images in the training data set as the Resnet50 deep learning model input, training the Resnet50 deep learning model to output probability values of the sample images and the pathological types corresponding to the sample images;
and adjusting parameters of the Resnet50 deep learning model according to the difference value of the output probability value and the probability value corresponding to the sample image in the training data set, continuing to train the Resnet50 deep learning model until the Resnet50 deep learning model converges, and taking the Resnet50 deep learning model at the time of convergence as the second labeling model.
6. The method for training a Chinese herbal medicine pest and disease identification model based on semi-supervised learning of claim 3, wherein the training of the Resnet50 deep learning model with the sample images in the training data set as the input of the Resnet50 deep learning model and the probability values of the sample images in the training data set and the corresponding pathological types as the output of the Resnet50 deep learning model comprises:
inputting the test data set into the trained second labeling model to judge whether the second labeling model outputs the test pathology type of the test data set and the test probability value corresponding to the test medical record type;
and if the test pathology type of the test data set and the test probability value corresponding to the test medical record type are output, the second labeling model is successfully trained.
7. The method for training a Chinese herbal medicine pest recognition model based on semi-supervised learning of claim 1, wherein the method further comprises:
and uploading the Chinese herbal medicine pest and disease damage identification model to a block chain.
8. The utility model provides a system based on semi-supervised learning training chinese herbal medicine plant diseases and insect pests recognition model which characterized in that includes:
the system comprises a first acquisition module, a second acquisition module and a third acquisition module, wherein the first acquisition module is used for acquiring an annotated image dataset which comprises a plurality of sample images, and each sample image is provided with a sample pathology type label corresponding to the sample image;
the first training module is used for training a Resnet50 deep learning model based on the sample images in the annotation image dataset to obtain a first annotation model;
the second acquisition module is used for acquiring an unlabeled image data set, inputting a plurality of unlabeled images contained in the unlabeled image set into the first labeling model, and obtaining a pathological type corresponding to each unlabeled image and a probability value corresponding to the pathological type;
the second training module is used for combining the labeled image data set, the pathological types corresponding to the unlabeled images and the probability values corresponding to the pathological types into a new training data set, and training the finetune model corresponding to the first labeled model based on the new training data set to obtain a Chinese herbal medicine pest and disease identification model;
the identification module is used for acquiring a Chinese herbal medicine image to be labeled and inputting the Chinese herbal medicine image to be labeled into the Chinese herbal medicine pest and disease identification model so as to obtain a target pathological type of the Chinese herbal medicine image and a target probability value corresponding to the target pathological type.
9. A computer device comprising a memory, a processor, a computer program stored on the memory and executable on the processor, the computer program when executed by the processor implementing the steps of the method for training a chinese herbal medicine pest identification model based on semi-supervised learning as claimed in any one of claims 1-7.
10. A computer-readable storage medium having stored therein a computer program executable by at least one processor to cause the at least one processor to perform the steps of the method for training a chinese herbal pest recognition model based on semi-supervised learning as claimed in any one of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011506545.1A CN112560964A (en) | 2020-12-18 | 2020-12-18 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011506545.1A CN112560964A (en) | 2020-12-18 | 2020-12-18 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112560964A true CN112560964A (en) | 2021-03-26 |
Family
ID=75030452
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011506545.1A Pending CN112560964A (en) | 2020-12-18 | 2020-12-18 | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112560964A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065609A (en) * | 2021-04-22 | 2021-07-02 | 平安国际智慧城市科技股份有限公司 | Image classification method and device, electronic equipment and readable storage medium |
| CN113344896A (en) * | 2021-06-24 | 2021-09-03 | 鹏城实验室 | Breast CT image focus segmentation model training method and system |
| CN113378835A (en) * | 2021-06-28 | 2021-09-10 | 北京百度网讯科技有限公司 | Labeling model training method, sample labeling method and related device |
| CN113420797A (en) * | 2021-06-08 | 2021-09-21 | 杭州知衣科技有限公司 | Online learning image attribute identification method and system |
| CN113506579A (en) * | 2021-05-27 | 2021-10-15 | 华南师范大学 | Insect pest recognition method based on artificial intelligence and sound and robot |
| CN113688959A (en) * | 2021-10-26 | 2021-11-23 | 寿光得峰生态农业有限公司 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
| CN114037868A (en) * | 2021-11-04 | 2022-02-11 | 杭州医策科技有限公司 | Image recognition model generation method and device |
| CN115082797A (en) * | 2022-07-05 | 2022-09-20 | 安徽农业大学 | Early wheat scab monitoring and early warning method and system in field environment |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108021926A (en) * | 2017-09-28 | 2018-05-11 | 东南大学 | A kind of vehicle scratch detection method and system based on panoramic looking-around system |
| CN108764281A (en) * | 2018-04-18 | 2018-11-06 | 华南理工大学 | A kind of image classification method learning across task depth network based on semi-supervised step certainly |
| CN109620223A (en) * | 2018-12-07 | 2019-04-16 | 北京工业大学 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
| CN110060244A (en) * | 2019-04-15 | 2019-07-26 | 深圳市麦迪普科技有限公司 | The system and method for cell detection and segmentation based on deep learning neural network |
| CN111046907A (en) * | 2019-11-02 | 2020-04-21 | 国网天津市电力公司 | Semi-supervised convolutional network embedding method based on multi-head attention mechanism |
| KR102093080B1 (en) * | 2019-12-06 | 2020-04-27 | 주식회사 애자일소다 | System and method for classifying base on generative adversarial network using labeled data and unlabled data |
| CN111141211A (en) * | 2019-12-24 | 2020-05-12 | 深圳信息职业技术学院 | Deep learning-based capacitance positioning detection method, server and detection system |
| CN111191732A (en) * | 2020-01-03 | 2020-05-22 | 天津大学 | Target detection method based on full-automatic learning |
| CN111325739A (en) * | 2020-02-28 | 2020-06-23 | 北京推想科技有限公司 | Method and device for detecting lung focus and training method of image detection model |
| CN111476256A (en) * | 2019-01-24 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Model training method and device based on semi-supervised learning and electronic equipment |
| CN111639540A (en) * | 2020-04-30 | 2020-09-08 | 中国海洋大学 | Semi-supervised character re-recognition method based on camera style and human body posture adaptation |
| CN111767849A (en) * | 2020-06-29 | 2020-10-13 | 京东数字科技控股有限公司 | Crop pest and disease identification method and device and storage medium |
| CN111834014A (en) * | 2020-07-17 | 2020-10-27 | 北京工业大学 | Medical field named entity identification method and system |
-
2020
- 2020-12-18 CN CN202011506545.1A patent/CN112560964A/en active Pending
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108021926A (en) * | 2017-09-28 | 2018-05-11 | 东南大学 | A kind of vehicle scratch detection method and system based on panoramic looking-around system |
| CN108764281A (en) * | 2018-04-18 | 2018-11-06 | 华南理工大学 | A kind of image classification method learning across task depth network based on semi-supervised step certainly |
| CN109620223A (en) * | 2018-12-07 | 2019-04-16 | 北京工业大学 | A kind of rehabilitation of stroke patients system brain-computer interface key technology method |
| CN111476256A (en) * | 2019-01-24 | 2020-07-31 | 北京京东尚科信息技术有限公司 | Model training method and device based on semi-supervised learning and electronic equipment |
| CN110060244A (en) * | 2019-04-15 | 2019-07-26 | 深圳市麦迪普科技有限公司 | The system and method for cell detection and segmentation based on deep learning neural network |
| CN111046907A (en) * | 2019-11-02 | 2020-04-21 | 国网天津市电力公司 | Semi-supervised convolutional network embedding method based on multi-head attention mechanism |
| KR102093080B1 (en) * | 2019-12-06 | 2020-04-27 | 주식회사 애자일소다 | System and method for classifying base on generative adversarial network using labeled data and unlabled data |
| CN111141211A (en) * | 2019-12-24 | 2020-05-12 | 深圳信息职业技术学院 | Deep learning-based capacitance positioning detection method, server and detection system |
| CN111191732A (en) * | 2020-01-03 | 2020-05-22 | 天津大学 | Target detection method based on full-automatic learning |
| CN111325739A (en) * | 2020-02-28 | 2020-06-23 | 北京推想科技有限公司 | Method and device for detecting lung focus and training method of image detection model |
| CN111639540A (en) * | 2020-04-30 | 2020-09-08 | 中国海洋大学 | Semi-supervised character re-recognition method based on camera style and human body posture adaptation |
| CN111767849A (en) * | 2020-06-29 | 2020-10-13 | 京东数字科技控股有限公司 | Crop pest and disease identification method and device and storage medium |
| CN111834014A (en) * | 2020-07-17 | 2020-10-27 | 北京工业大学 | Medical field named entity identification method and system |
Non-Patent Citations (3)
| Title |
|---|
| DONG-HYUN LEE等: "Pseudo-Label : The Simple and E cient Semi-Supervised Learning Method for Deep Neural Networks", 《COMPUTER SCIENCE》, pages 1 - 7 * |
| 刘坤等: "基于半监督生成对抗网络Ｘ光图像分类算法", 《光学学报》, vol. 39, no. 8, pages 1 - 8 * |
| 姚佳奇等: "GCN-PU：基于图卷积网络的PU 文本分类算法", 《计算机工程与应用》, pages 1 - 9 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113065609A (en) * | 2021-04-22 | 2021-07-02 | 平安国际智慧城市科技股份有限公司 | Image classification method and device, electronic equipment and readable storage medium |
| CN113065609B (en) * | 2021-04-22 | 2024-04-09 | 深圳赛安特技术服务有限公司 | Image classification method, device, electronic equipment and readable storage medium |
| CN113506579B (en) * | 2021-05-27 | 2024-01-23 | 华南师范大学 | Insect pest identification method and robot based on artificial intelligence and voice |
| CN113506579A (en) * | 2021-05-27 | 2021-10-15 | 华南师范大学 | Insect pest recognition method based on artificial intelligence and sound and robot |
| CN113420797A (en) * | 2021-06-08 | 2021-09-21 | 杭州知衣科技有限公司 | Online learning image attribute identification method and system |
| CN113344896A (en) * | 2021-06-24 | 2021-09-03 | 鹏城实验室 | Breast CT image focus segmentation model training method and system |
| CN113378835A (en) * | 2021-06-28 | 2021-09-10 | 北京百度网讯科技有限公司 | Labeling model training method, sample labeling method and related device |
| CN113688959A (en) * | 2021-10-26 | 2021-11-23 | 寿光得峰生态农业有限公司 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
| CN113688959B (en) * | 2021-10-26 | 2022-02-18 | 寿光得峰生态农业有限公司 | Plant disease and insect pest diagnosis method and system based on artificial intelligence |
| CN114037868A (en) * | 2021-11-04 | 2022-02-11 | 杭州医策科技有限公司 | Image recognition model generation method and device |
| CN114037868B (en) * | 2021-11-04 | 2022-07-01 | 杭州医策科技有限公司 | Image recognition model generation method and device |
| CN115082797B (en) * | 2022-07-05 | 2023-08-22 | 安徽农业大学 | Early wheat scab monitoring and early warning method and system in field environment |
| CN115082797A (en) * | 2022-07-05 | 2022-09-20 | 安徽农业大学 | Early wheat scab monitoring and early warning method and system in field environment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112560964A (en) | Method and system for training Chinese herbal medicine pest and disease identification model based on semi-supervised learning | |
| CN109815339B (en) | Knowledge extraction method and device based on TextCNN, computer equipment and storage medium | |
| CN110688288B (en) | Automatic test method, device, equipment and storage medium based on artificial intelligence | |
| CN110532884B (en) | Pedestrian re-recognition method, device and computer readable storage medium | |
| CN109271958B (en) | Face age identification method and device | |
| JP7182021B2 (en) | KEYPOINT DETECTION METHOD, KEYPOINT DETECTION DEVICE, ELECTRONIC DEVICE, AND STORAGE MEDIUM | |
| CN111767707A (en) | Method, device, equipment and storage medium for detecting Rayleigh case | |
| CN111414353A (en) | Intelligent missing data filling method and device and computer readable storage medium | |
| CN110866115B (en) | Sequence labeling method, system, computer equipment and computer readable storage medium | |
| WO2021027142A1 (en) | Picture classification model training method and system, and computer device | |
| CN111401339B (en) | Method and device for identifying age of person in face image and electronic equipment | |
| CN111832581B (en) | Lung feature recognition method and device, computer equipment and storage medium | |
| CN109918498B (en) | Problem warehousing method and device | |
| CN109271957B (en) | Face gender identification method and device | |
| CN110782996A (en) | Construction method and device of medical database, computer equipment and storage medium | |
| CN111931002A (en) | Matching method and related equipment | |
| CN110502620B (en) | Method, system and computer equipment for generating guide diagnosis similar problem pairs | |
| CN112966685A (en) | Attack network training method and device for scene text recognition and related equipment | |
| CN110162957B (en) | Authentication method and device for intelligent equipment, storage medium and electronic device | |
| CN114387656A (en) | Face changing method, device, equipment and storage medium based on artificial intelligence | |
| CN112818946A (en) | Training of age identification model, age identification method and device and electronic equipment | |
| CN111666991A (en) | Convolutional neural network-based pattern recognition method and device and computer equipment | |
| CN110750637B (en) | Text abstract extraction method, device, computer equipment and storage medium | |
| CN112132269A (en) | Model processing method, device, equipment and storage medium | |
| CN111967383A (en) | Age estimation method, and training method and device of age estimation model |
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 |