CN112215082A - Plant leaf image identification method - Google Patents
Plant leaf image identification method Download PDFInfo
- Publication number
- CN112215082A CN112215082A CN202010978200.XA CN202010978200A CN112215082A CN 112215082 A CN112215082 A CN 112215082A CN 202010978200 A CN202010978200 A CN 202010978200A CN 112215082 A CN112215082 A CN 112215082A
- Authority
- CN
- China
- Prior art keywords
- image data
- leaf image
- leaf
- plants
- matrix decomposition
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- 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
- G06V10/30—Noise filtering
Abstract
The invention discloses a plant leaf image identification method, which comprises the following steps: acquiring leaf image data sets of a plurality of plants; wherein the leaf image dataset of each plant contains 60-80 leaf images; carrying out normalization pretreatment on leaf image data contained in the leaf image data sets of the plants; carrying out depth matrix decomposition on the normalized and preprocessed leaf image data to retain target characteristic information of the leaf image data; inputting the leaf image data subjected to the depth matrix decomposition into a convolutional neural network for learning training to obtain a depth learning model; and inputting the leaf images of the plants to be identified into the deep learning model, and identifying the types of the plants to be identified. According to the invention, the data complexity can be effectively reduced and the characteristic information can be reserved in the plant leaf identification, and the plant leaf image identification can be efficiently realized.
Description
Technical Field
The invention relates to the field of plant leaf image identification, in particular to a plant leaf image identification method.
Background
There are about 40 thousands of plants in nature, of which over 25 thousands have been named and their information structures entered into databases, and plants serve as important roles in human survival activities, such as the fields of agriculture, environmentality, medicine, and the like.
The classification and identification of plants usually adopt a mode of selecting local characteristics (such as leaves, flowers, fruits, stems, branches and the like of the plants) of the plants for classification and identification, and because the leaves of the plants have longer survival time and are convenient to collect compared with other organs, the leaves of the plants are often used as the identification characteristics of the plants and the main reference organs for recognizing the plants; in addition, leaf shape, texture, vein and the like are important indexes for studying morphological variation and differentiation of plant species, and therefore, plant identification using leaves as entry points is the most direct, effective and simple method.
The traditional plant identification is a manual identification method, along with the development of computer technology and pattern recognition algorithm, the method for identifying plant types based on images replaces the manual identification method, the existing image identification method is characterized in that corresponding features are extracted from leaf images through related algorithms, then the extracted features are classified, and then classification of corresponding plants is realized. With the development and application of deep learning, the convolutional neural network has a prominent effect on image recognition, in the application of plant leaf recognition, the convolutional neural network does not need to be artificially divided to extract features, the model directly performs learning training on leaf image data and extracts the features, and experimental results show that the accuracy and the stability of the method are remarkably improved. However, the deep learning model needs large data support, and it is very important to select a proper model and data, and for plant leaf data, the model is usually high-dimensional and complex, and has many characteristic parameters, which easily results in phenomena of long training time, unstable accuracy, overfitting, etc. of the deep learning network, so the effect of the convolutional neural network for identifying images depends on the training of the deep learning network, i.e. the processing of plant leaf data, and therefore, a plant leaf image identification method capable of effectively reducing the data complexity and retaining the characteristic information in plant leaf identification needs to be designed, and the plant leaf image identification can be efficiently realized.
Disclosure of Invention
The invention provides a plant leaf image identification method, which can effectively reduce the data complexity and retain the characteristic information in the plant leaf identification, and can efficiently realize the plant leaf image identification.
The invention provides a plant leaf image identification method, which comprises the following steps:
acquiring leaf image data sets of a plurality of plants; wherein the leaf image dataset of each plant contains 60-80 leaf images;
carrying out normalization pretreatment on leaf image data contained in the leaf image data sets of the plants;
carrying out depth matrix decomposition on the normalized and preprocessed leaf image data to retain target characteristic information of the leaf image data;
inputting the leaf image data subjected to the depth matrix decomposition into a convolutional neural network for learning training to obtain a deep learning model;
and inputting the leaf images of the plants to be identified into the deep learning model, and identifying the types of the plants to be identified.
Preferably, the normalization preprocessing is performed on leaf image data contained in the leaf image data sets of the several plants, and includes the following steps:
and adopting Python to process the jpg format, graying and pixel value unification of the leaf image data, and dividing the processed leaf image data into a training set and a verification set, wherein the proportion of the training set to the verification set is 8: 2.
Preferably, after the leaf image data included in the leaf image data sets of the several plants are all subjected to normalization preprocessing, the method further includes the following steps:
denoising the blade image data after the normalization preprocessing through a self-adaptive median filtering algorithm;
performing depth matrix decomposition on the normalized and preprocessed leaf image data to retain target characteristic information of the leaf image data, specifically:
and carrying out depth matrix decomposition on the denoised blade image data to retain target characteristic information of the blade image data.
Preferably, the algorithm flow of the adaptive median filtering algorithm denoising comprises two of a and B, wherein,
A:A1=Zmed-Zmin;A2=Zmed-Zmax;
If A1>0,and A2<0,then run to B;
Else increase the window size;
If the increased size is≤Smax,repeat A;
Otherwise,directly output Zmed;
B:B1=Zxy-Zmin;B2=Zxy-Zmax;
If B1>0,and B2<0,then outputZxy;
Else outputZmed;
wherein S isxyIs the action area of the filter, namely the area covered by the filter window, the central point of the area is the y-th row and x-th column pixel points in the image, ZminIs SxyMiddle minimum gray value, ZmaxIs SxyMiddle maximum gray value, ZmedIs SxyMedian of all gray values in, ZxyExpressing the gray value of the pixel point of the y row and the x column in the image, SmaxIs SxyThe maximum window size allowed.
Preferably, the depth matrix decomposition model is as follows:wherein m hidden layers are decomposed:hidden layerIs nonnegativeEach layer is also suitable for cluster interpretation, in leaf image data,namely the shape, texture and venation feature information in the clustering image.
Preferably, the convolutional neural network model is as follows: the model is composed of 10 layers, the input image data is X × X, the activation function is "relu", and the model structure can be used to input (X, X, 1) -C1(20, 5, 1) -P1(2, 2, "same") -C2(20, 5, 1) -P2(2, 2, "same") -C3(20, 5, 1) -P3(2, 2, "same") -fc (1024) -fc (786) -fc (128) -fc (8), wherein C (20, 5, 1) is a convolution kernel of each channel in a convolution process, the step size is 1, and the boundary of the feature map is filled by adopting a "0" filling mode.
Preferably, fc generates output vectors for the fully-connected layers, each containing a corresponding unit, and Dropout (0, 5) is added at both layers fc (786) -fc (128) to prevent overfitting.
Compared with the prior art, the invention has the following beneficial effects:
according to the method, the depth matrix decomposition and the convolutional neural network are combined, the image identification process can be simplified through the convolutional neural network, the data are input into the deep learning neural network model, the identification result can be directly obtained, and the identification efficiency and the accuracy are improved; through the depth matrix decomposition, the problems of long network training time, unstable accuracy, overfitting and the like caused by high-dimensional complexity and more characteristic parameters of leaf image data can be effectively solved, the complexity of the data can be effectively reduced, the characteristic information can be retained, and the plant leaf image recognition can be efficiently realized.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:
FIG. 1 is a flow chart of a plant leaf image recognition method according to an embodiment of the present invention;
FIG. 2 is a comparison graph of the adaptive median filtering effect according to the first embodiment of the present invention;
FIG. 3 is a schematic diagram of a depth matrix decomposition according to a first embodiment of the present invention;
FIG. 4 is a comparison graph of the decomposition effect of the depth matrix according to the first embodiment of the present invention;
FIG. 5 is a diagram of a convolutional neural network according to a first embodiment of the present invention;
FIG. 6 is a flow chart of an adaptive median filtering according to a first embodiment of the present invention;
FIG. 7 is a diagram of a plant sample according to a first embodiment of the present invention.
Detailed Description
In order to make the technical scheme and implementation method of the present invention clearer, the following describes the implementation process in detail with reference to the drawings and preferred embodiments of the present invention. The embodiments described are only a part of the embodiments of the present invention, and all other embodiments obtained by those skilled in the art without any inventive work are within the scope of the present invention.
Example one
An embodiment of the present invention provides a plant leaf image identification method, and fig. 1 is a flowchart of a plant leaf image identification method according to an embodiment of the present invention, as shown in fig. 1, including the following steps:
step S101: acquiring leaf image data sets of a plurality of plants;
in an embodiment of the present invention, the leaf image dataset of each of the plurality of plants comprises 60-80 leaf images; FIG. 7 is a sample diagram of a leaf of a part of a plant obtained according to the present embodiment;
step S102: carrying out normalization pretreatment on leaf image data contained in leaf image data sets of a plurality of plants;
in the embodiment of the present invention, the specific implementation manner of step S102 is to adopt Python to process the jpg format, graying, and pixel value of the leaf image data to make them uniform, and divide the processed leaf image data into a training set and a verification set, where the ratio of the training set to the verification set is 8: 2;
step S103: denoising the blade image data after the normalization preprocessing through a self-adaptive median filtering algorithm;
in the embodiment of the present invention, the basic principle of median filtering is as follows: the method is characterized in that a two-dimensional sliding template with a certain structure is used, the size of the pixel values in the template is sequenced to generate a two-dimensional data sequence which is monotonically rising or monotonically falling, and the two-dimensional median is filtered and output as follows: g (x, y) ═ med { f (x-k, y-l), (k, l ∈ w) }, where f (x, y), g (x, y) are the original image and the processed image respectively, w is a two-dimensional template, typically a 2 × 2, 3 × 3 region, and may also be different shapes, such as a line, a circle, a cross, a circular ring, etc.; as shown in fig. 2, which is a comparison graph of the effect of grayed plant image data after median filtering in windows of different sizes, it can be seen that the image after median filtering becomes smooth, but the image is blurred due to oversize of the window and is not beneficial to recognition and classification, so that the problem can be solved well by adopting adaptive median filtering in the invention, and the integrity of the image is maintained;
in the embodiment of the invention, the denoising algorithm flow of the adaptive median filtering algorithm comprises two types A and B, wherein,
A:A1=Zmed-Zmin;A2=Zmed-Zmax;
If A1>0,and A2<0,then run to B;
Else increase the window size;
If the increased size is≤Smax,repeat A;
Otherwise,directly outputZmed;
B:B1=Zxy-Zmin;B2=Zxy-Zmax;
If B1>0,and B2<0,then outputZxy;
Else outputZmed;
wherein S isxyIs the action area of the filter, namely the area covered by the filter window, the central point of the area is the y-th row and x-th column pixel points in the image, ZminIs SxyMiddle minimum gray value, ZmaxIs SxyMiddle maximum gray value, ZmedIs SxyMedian of all gray values in, ZxyExpressing the gray value of the pixel point of the y row and the x column in the image, SmaxIs SxyThe maximum window size allowed; the adaptive median filtering flow is shown in fig. 6;
step S104: carrying out depth matrix decomposition on the denoised blade image data to retain target characteristic information of the blade image data;
in the embodiment of the invention, the depth matrix decomposition is carried out on the denoised blade image data through the step S104, so that the main characteristic information of the blade image data can be reserved, the data complexity is reduced, the parameters are reduced for the subsequent network training, and the efficiency and the stability are improved; the schematic diagram of the depth matrix decomposition is shown in fig. 3, the effect diagram of the depth matrix decomposition is shown in fig. 4, and as can be seen from fig. 4, along with the reduction of the characteristic parameters, the efficiency is improved, but the accuracy is reduced, so that the excessive removal of the characteristic parameters is not beneficial to the identification of the convolutional neural network, and therefore, the main characteristic information of the leaf image data is reserved through the depth matrix decomposition, and the efficiency and the accuracy are improved at the same time;
in the embodiment of the present invention, the depth matrix decomposition model is as follows:wherein m hidden layers are decomposed:hidden layerNon-negative, each layer is also suitable for cluster interpretation, in leaf image data,the information is the characteristic information of shape, texture and venation in the clustering image;
further, in the training of the depth matrix decomposition model, the initialization data matrix X is approximately equal to Z1H1WhereinThe feature matrix H is then decomposed1≈Z2H2WhereinAnd sequentially training the rule to the last layer in a circulating way, alternately optimizing two factors, and in order to reduce the model reconstruction error, the objective function is as the formula:updating the weight matrix Z by keeping ithThe remainder of the layer weights, then for ZiTake the minimum objective function such thatThe update rule is as follows:wherein psi ═ Z1...Zi-1And' represents a pseudo-inverse matrix,is the iththReconstruction matrix of feature matrix, HiThe update rule is as follows:in the process of identifying the plant leaf images, the images contain a large amount of plant phenotype characteristic information, the traditional matrix decomposition is difficult to obtain internal deep layer information, and the deep semi-NMF carries out characteristic momentThe deep decomposition of the array can not only carry out hierarchical characteristic expression on the data, but also obtain good clustering performance;
step S105: inputting the leaf image data subjected to the depth matrix decomposition into a convolutional neural network for learning training to obtain a deep learning model;
in the embodiment of the present invention, the structure of the convolutional neural network is shown in fig. 5, and the convolutional neural network model is as follows: the model is composed of 10 layers, the input image data is X × X, the activation function is "relu", the model structure can be used for inputting (X, X, 1) -C1(20, 5, 1) -P1(2, 2, "same") -C2(20, 5, 1) -P2(2, 2, "same") -C3(20, 5, 1) -P3(2, 2, "same") -fc (1024) -fc (786) -fc (128) -fc (8), wherein C (20, 5, 1) is a convolution kernel of each channel in a convolution process, the step size is 1, and the boundary of the feature map is paved in a 0-filling mode; all pooling layers (2 multiplied by 2) adopt a maximum pooling mode, so that distortion can be effectively avoided;
fc generates output vectors for the fully-connected layers, each output vector contains a corresponding unit, and Dropout (0, 5) is added in two layers of fc (786) -fc (128) to prevent overfitting; and finally, estimating each image by adopting a softmax regression function to express the probability of class membership. The model training process may set the epochs to 20 and the batch size to 30, and an Adam optimizer is used to set the learning rate to 0.0001;
step S106: and inputting the leaf images of the plants to be identified into the deep learning model, and identifying the types of the plants to be identified.
In summary, according to the embodiments, the depth matrix decomposition is combined with the convolutional neural network, the image recognition process can be simplified through the convolutional neural network, the data is input into the deep learning neural network model to directly obtain the recognition result, and the recognition efficiency and the accuracy are improved; through the depth matrix decomposition, the problems of long network training time, unstable accuracy, overfitting and the like caused by high-dimensional complexity and more characteristic parameters of leaf image data can be effectively solved, the complexity of the data can be effectively reduced, the characteristic information can be retained, and the plant leaf image recognition can be efficiently realized.
Claims (7)
1. A plant leaf image identification method is characterized by comprising the following steps:
acquiring leaf image data sets of a plurality of plants; wherein the leaf image dataset of each plant contains 60-80 leaf images;
carrying out normalization pretreatment on leaf image data contained in the leaf image data sets of the plants;
carrying out depth matrix decomposition on the normalized and preprocessed leaf image data to retain target characteristic information of the leaf image data;
inputting the leaf image data subjected to the depth matrix decomposition into a convolutional neural network for learning training to obtain a deep learning model;
and inputting the leaf images of the plants to be identified into the deep learning model, and identifying the types of the plants to be identified.
2. The method according to claim 1, wherein the leaf image data included in the leaf image data sets of the plurality of plants are each subjected to a normalization preprocessing, comprising the steps of:
and adopting Python to process the jpg format, graying and pixel value unification of the leaf image data, and dividing the processed leaf image data into a training set and a verification set, wherein the proportion of the training set to the verification set is 8: 2.
3. The method according to claim 2, wherein after the normalization preprocessing of the leaf image data included in the leaf image data sets of the plurality of plants, the method further comprises the following steps:
denoising the blade image data after the normalization preprocessing through a self-adaptive median filtering algorithm;
performing depth matrix decomposition on the normalized and preprocessed leaf image data to retain target characteristic information of the leaf image data, specifically:
and carrying out depth matrix decomposition on the denoised blade image data to retain target characteristic information of the blade image data.
4. The method of claim 3, wherein the algorithm flow of the adaptive median filtering algorithm denoising comprises A and B, wherein,
A:A1=Zmed-Zmin;A2=Zmed-Zmax;
If A1>0,and A2<0,then run to B;
Else increase the window size;
If the increased size is≤Smax,repeat A;
Otherwise,directly output Zmed;
B:B1=Zxy-Zmin;B2=Zxy-Zmax;
If B1>0,and B2<0,then output Zxy;
Else output Zmed;
wherein S isxyIs the action area of the filter, namely the area covered by the filter window, the central point of the area is the y-th row and x-th column pixel points in the image, ZminIs SxyMiddle minimum gray value, ZmaxIs SxyMiddle maximum gray value, ZmedIs SxyMedian of all gray values in, ZxY tableShowing the gray value of the pixel point of the y row and x column in the image, SmaxIs SxyThe maximum window size allowed.
5. The method of claim 4, wherein the depth matrix decomposition model is as follows:wherein m hidden layers are decomposed:hidden layerNon-negative, each layer is also suitable for cluster interpretation, in leaf image data,namely the shape, texture and venation feature information in the clustering image.
6. The method of claim 5, wherein the convolutional neural network model is as follows: the model is composed of 10 layers, the input image data is X × X, the activation function is "relu", and the model structure can be used to input (X, X, 1) -C1(20, 5, 1) -P1(2, 2, "same") -C2(20, 5, 1) -P2(2, 2, "same") -C3(20, 5, 1) -P3(2, 2, "same") -fc (1024) -fc (786) -fc (128) -fc (8), wherein C (20, 5, 1) is a convolution kernel of each channel in a convolution process, the step size is 1, and the boundary of the feature map is filled by adopting a "0" filling mode.
7. Method according to claim 6, characterized in that fc generates output vectors for fully connected layers, each output vector containing the corresponding units, and Dropout (0, 5) is added at both layers fc (786) -fc (128) to prevent overfitting.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010978200.XA CN112215082A (en) | 2020-09-17 | 2020-09-17 | Plant leaf image identification method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010978200.XA CN112215082A (en) | 2020-09-17 | 2020-09-17 | Plant leaf image identification method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112215082A true CN112215082A (en) | 2021-01-12 |
Family
ID=74050488
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010978200.XA Pending CN112215082A (en) | 2020-09-17 | 2020-09-17 | Plant leaf image identification method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112215082A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113361589A (en) * | 2021-06-01 | 2021-09-07 | 杨晶晶 | Rare or endangered plant leaf identification method based on transfer learning and knowledge distillation |
CN114027052A (en) * | 2021-10-20 | 2022-02-11 | 华南农业大学 | Illumination regulation and control system for plant reproductive development |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109325515A (en) * | 2018-08-10 | 2019-02-12 | 江苏理工学院 | Matrix of depths decomposition method and image clustering method based on part study regularization |
CN109522924A (en) * | 2018-09-28 | 2019-03-26 | 浙江农林大学 | A kind of broad-leaf forest wood recognition method based on single photo |
CN110942108A (en) * | 2019-12-13 | 2020-03-31 | 深圳大学 | Face image clustering method and device and computer readable storage medium |
CN111476280A (en) * | 2020-03-27 | 2020-07-31 | 海南医学院 | Plant leaf identification method and system |
-
2020
- 2020-09-17 CN CN202010978200.XA patent/CN112215082A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109325515A (en) * | 2018-08-10 | 2019-02-12 | 江苏理工学院 | Matrix of depths decomposition method and image clustering method based on part study regularization |
CN109522924A (en) * | 2018-09-28 | 2019-03-26 | 浙江农林大学 | A kind of broad-leaf forest wood recognition method based on single photo |
CN110942108A (en) * | 2019-12-13 | 2020-03-31 | 深圳大学 | Face image clustering method and device and computer readable storage medium |
CN111476280A (en) * | 2020-03-27 | 2020-07-31 | 海南医学院 | Plant leaf identification method and system |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113361589A (en) * | 2021-06-01 | 2021-09-07 | 杨晶晶 | Rare or endangered plant leaf identification method based on transfer learning and knowledge distillation |
CN114027052A (en) * | 2021-10-20 | 2022-02-11 | 华南农业大学 | Illumination regulation and control system for plant reproductive development |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110929029A (en) | Text classification method and system based on graph convolution neural network | |
CN105844285A (en) | Cucumber disease identification method and apparatus based on image information | |
CN111000553B (en) | Intelligent classification method for electrocardiogram data based on voting ensemble learning | |
CN108648191A (en) | Pest image-recognizing method based on Bayes's width residual error neural network | |
CN111931914A (en) | Convolutional neural network channel pruning method based on model fine tuning | |
Reddy et al. | Resnet-based modified red deer optimization with DLCNN classifier for plant disease identification and classification | |
CN105512681A (en) | Method and system for acquiring target category picture | |
CN112215082A (en) | Plant leaf image identification method | |
Yu et al. | A recognition method of soybean leaf diseases based on an improved deep learning model | |
CN111597907B (en) | Anti-noise meta-learning-based face recognition method and system | |
CN109242106A (en) | sample processing method, device, equipment and storage medium | |
CN114781441A (en) | EEG motor imagery classification method and multi-space convolution neural network model | |
Shankar et al. | A comprehensive analysis of deep learning techniques for recognition of flower species | |
CN114049935A (en) | HER2 image classification system based on multi-convolution neural network | |
Padole et al. | Graph wavelet-based multilevel graph coarsening and its application in graph-CNN for alzheimer’s disease detection | |
CN113221913A (en) | Agriculture and forestry disease and pest fine-grained identification method and device based on Gaussian probability decision-level fusion | |
Wang et al. | Generalized convolutional sparse coding with unknown noise | |
CN109934281B (en) | Unsupervised training method of two-class network | |
CN115205580A (en) | Multi-branch residual convolution neural network model and image classification method thereof | |
CN111178174B (en) | Urine formed component image identification method based on deep convolutional neural network | |
CN114548291A (en) | Ecological biological identification method based on MR-CNN algorithm | |
CN114494828A (en) | Grape disease identification method and device, electronic equipment and storage medium | |
CN111882441A (en) | User prediction interpretation Treeshap method based on financial product recommendation scene | |
CN113469129A (en) | Wheat impurity image detection method based on convolutional neural network | |
CN115601634A (en) | Image blade identification method and device based on hierarchical attention mechanism |
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 |