CN109446985A - Multi-angle plants identification method based on vector neural network - Google Patents
Multi-angle plants identification method based on vector neural network Download PDFInfo
- Publication number
- CN109446985A CN109446985A CN201811263198.7A CN201811263198A CN109446985A CN 109446985 A CN109446985 A CN 109446985A CN 201811263198 A CN201811263198 A CN 201811263198A CN 109446985 A CN109446985 A CN 109446985A
- Authority
- CN
- China
- Prior art keywords
- vector
- plant
- neural network
- angle
- image
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the multi-angle plants identification methods based on vector neural network, this method mainly comprises the steps of: the floristics identified according to actual needs, the plant image of certain amount multi-angle of view is collected for every class plant, it is spare to form data set after manual sort label;By the plant image input vector neural network in data set, the training for having supervision is carried out to network, allowing network that can automatically extract the information with spatial variations robustness from multiple angular images of kindred plant indicates;Feature is sent into feature classifiers, the highest classification of classification results probability is the plant generic.Vector neural network used in the present invention, pole drastically reduce dependence of the traditional neural network to data volume, just can construct and train the neural network with stronger plant characteristics abstracting power with the multi-angle of view plant image of relatively small amount, and model identifies floristics.
Description
Technical field
The present invention relates to images to judge field, in particular to the multi-angle plants identification method based on vector neural network.
Background technique
Plant and the mankind are closely bound up, closely coupled over the past thousands of years.With the progress of human civilization, plant is but by more
Carry out more serious destruction.Therefore, identification classification is carried out to plant, establishes plant digitalization resource library, to help the protection of plant,
Have great importance.And plant leaf digital image machine recognition algorithm can undoubtedly greatly speed up the classification work of plant.
In classical pattern-recognition, the feature (such as SIFT, HOG, LBP feature) of preset plant image is usually extracted in advance.It mentions
After taking feature, feature is encoded, such as common BoW, FisherVector etc..Then feature is put into a classifier,
Such as SVM, 2 classification are carried out, optimal classification surface is trained, finds the feature that can most represent certain class plant, are removed unrelated to classifying
With autocorrelative feature.However, the extraction of these features too relies on the experience and subjective consciousness of people, the feature extracted is not
It is very big with being influenced on classification performance, or even the sequence of feature extracted also will affect last plant classification performance.Meanwhile image
Pretreated quality also influences whether the feature extracted.
Summary of the invention
It is an object of the invention to: the multi-angle plants identification method based on vector neural network is provided, solves sky
Between on Feature Semantics related question, greatly save data collection cost, improve recognition accuracy and the scope of application.
The technical solution adopted by the invention is as follows:
Multi-angle plants identification method based on vector neural network, including, it is characterised in that: the method includes as follows
Step: S1: collecting plant multi-angle image, passes through the quasi- multiple angles for identifying plant Various Seasonal of image acquisition device
Organ topography, collect and arrange image resource by plant classification carry out taxonomic revision classification classify it is spare;S2: image
Training set production, by the way that the plant image for collecting acquisition in step S1 to be labeled as sub-category, and by the figure after mark
As data are uniformly adjusted to size of the same size;S3: identification vector neural network model training, by will be in step S2
Image data that treated is input in vector neural network, allows the network to extract plant image after successive ignition
Multidimensional characteristic, and feature switched into vector value by vector algorithm, wherein the long possibility for indicating its generic of vector field homoemorphism
Property, the direction of vector indicates the class instance parameter of plant;S4: vector neural network that the training in step S3 obtains is known
Other floristics, belongs to the image for the plant to be identified, the feature with unchanged view angle is independently extracted through network, generates most
Whole feature vector, and Classification Loss is calculated according to this vector field homoemorphism length, identification classification results name is obtained after classified calculating
Claim;S5: according to the classification results title calculated in step S4, searching for database or Internet resources data, and after showing search
Data details.
Further, the topography acquired in the step S1 is flower, leaf, fruit and the apparent plant of feature of plant
Organic image.
Further, the taxonomic revision in the step S1 is to be put into the plant data of acquisition together according to same type data
In one file.
Further, it is labeled as the file after the taxonomic revision in step S1 being numbered in the step S2, and
Botanical name corresponding to each number is listed in the form of text, and foring one can be used for training the vector neural network
Data set.
Further, the picture size in the step S2 is 128*128.
Further, the multidimensional characteristic in the step S3 is to extract after several master vector units, is had more
The feature of a dimension.
Further, the routing algorithm between vector of the vector algorithm in the step S3.
Further, the substance parameter in the step S3 has blade texture, vein distribution, shape information, substance parameter
Feature extraction is carried out by machine autonomous learning.
In conclusion by adopting the above-described technical solution, the beneficial effects of the present invention are:
1. establishing big-sample data library the present invention is based on the multi-angle plants identification method of vector neural network, drawing simultaneously
Enter invalid value option, using artificial intelligent depth learning model, learns and extract characteristic information automatically from sample database image, from
And floristics described in sample database image can be analyzed.
2., will with the incoherent picture of training image the present invention is based on the multi-angle plants identification method of vector neural network
It is automatically recognized as invalid value, the picture diversity for being actually subjected to identification is higher
3. diversity sample database is established, in difference the present invention is based on the multi-angle plants identification method of vector neural network
Under the conditions of acquire picture, all take as long as meeting basic demand, to improve anti-interference ability, when to prevent clinically applying
All kinds of interference are encountered to occur
Specific embodiment
All features disclosed in this specification or disclosed all methods or in the process the step of, in addition to mutually exclusive
Feature and/or step other than, can combine in any way.
Embodiment 1
Multi-angle plants identification method based on vector neural network, including, it is characterised in that: the method includes as follows
Step: S1: collecting plant multi-angle image, passes through the quasi- multiple angles for identifying plant Various Seasonal of image acquisition device
Organ topography, collect and arrange image resource by plant classification carry out taxonomic revision classification classify it is spare;S2: image
Training set production, by the way that the plant image for collecting acquisition in step S1 to be labeled as sub-category, and by the figure after mark
As data are uniformly adjusted to size of the same size;S3: identification vector neural network model training, by will be in step S2
Image data that treated is input in vector neural network, allows the network to extract plant image after successive ignition
Multidimensional characteristic, and feature switched into vector value by vector algorithm, wherein the long possibility for indicating its generic of vector field homoemorphism
Property, the direction of vector indicates the class instance parameter of plant;S4: vector neural network that the training in step S3 obtains is known
Other floristics, belongs to the image for the plant to be identified, the feature with unchanged view angle is independently extracted through network, generates most
Whole feature vector, and Classification Loss is calculated according to this vector field homoemorphism length, identification classification results name is obtained after classified calculating
Claim;S5: according to the classification results title calculated in step S4, searching for database or Internet resources data, and after showing search
Data details.
The working principle of the invention/the course of work are as follows: pass through the quasi- identification plant of image acquisition device different seasons first
The organ topography of multiple angles of section collects and arranges image resource and divided by plant classification progress taxonomic revision classification
Class is spare, is then carrying out training set of images production, by will be collected in step S1 the plant image of acquisition by it is sub-category into
Rower note, and the image data after mark is uniformly adjusted to size of the same size, then training image inputs, and converts pixel
The input net that feature (determines treated data) per batch of size according to the size of specific computer video memory in batch one by one
Network, the pixel convolution unit in network carry out convolution operation to original input picture.This pixel convolution unit is by 256 9*
9 convolution kernel composition, each convolution kernel is with the impression region of 9*9, the interval convolved image that step-length is 1, then often by one
ReLU activation primitive carries out a nonlinear transformation, and pixel intensity is just converted into multiple local features (here in this way
Have 256 convolution kernels, just can generate 256 basic local features), multiple local features obtained in pixel convolution unit are sent into
To master vector unit, the semantic feature that can be used for classifying further is extracted in this unit.This element is by multichannel
Convolution operation is composed, 32 channels, and there are one 8 dimension convolution, convolution kernel 9*9, step-length 2 in each channel.To feeding
Characteristic value further use convolution operation, multiple output vectors can be obtained after convolution.In order to preferably embody the table of vector
Danone power does normalization compression processing for the vector of output, so that short vector is compressed to almost nil, all length is all
No more than 1.As formula one wherein Vj be the j unit output vector, Sj is fully entering for it, formula one are as follows:
It obtains after normalized output vector to next layer, next layer is master vector unit or characteristic vector unit, is used
Feature locations correlation transmission method carrys out conduction parameter, specific as follows: such as formula two, in addition to the pixel convolution unit of first layer,
Fully entering for master vector Sj is to low one layer of output vector uj | the weighted sum of i.These vectors are all by low one layer of unit
It generates, is got by the output ui of unit and a weight matrix Wij multiplication.Wherein cijIt is the pass determined by feature conductive process
Contact number.Unit i and thereon in one layer the incidence coefficient of all units and be 1, this coefficient is bigger, the association between explanation
It spends bigger.Formula two are as follows:
Incidence coefficient cijIt is determined by formula three, wherein bijInitial value be 0, all can be by two features before each calculate
Dot product is added up between relating value, that is, vector between vector, i.e. bij=bij+vj·ui|j, incidence coefficient c has just been obtained in this wayij。
The vector extracted in master vector unit is passed to characteristic vector unit, this is a full connection unit, in this list
Be in member several corresponding with class categories 16 dimensional vectors (such as to identify 100 kinds of plants, be just arranged to here 100 to
Amount), determine according to vector field homoemorphism is long as a result, and being compared with the true value of label, the calculating loss (difference between true value and predicted value
Value), calculation formula such as formula 4:
Lk=Tkmax(0,m+-‖Vk‖)2+λ(1-Tk)max(0,‖Vk‖-m-)2(formula four)
Wherein, when predicted value is consistent with true mark value, Tk=1, conversely, Tk=0.According to classification actual classification quantity and
Training effect adjusts m+、m-With tri- parameters of λ, the present invention in recommend three parameters be once 0.9,0.1,0.5.
In present specification, vector neural network is also related to, for the extraction of the base pixel feature to input picture,
The vector neural network is made of a pixel convolution unit, a characteristic vector unit and several master vector units.Pixel
Convolution unit is responsible for receiving original images to be recognized, and master vector unit reception pixel convolution unit is defeated with other master vector units
Outgoing vector.Multitiered network structure is extracted since the base pixel feature of image, is then further extracted in higher
Semantic feature, convolution kernel can as needed customized adjusting, export as vector.Master vector unit: the feature from bottom is received
Vector is further abstracted advanced features, obtains the high-level semantics feature of plant.In the convolution operation that this element is by multichannel
It is composed.Characteristic vector unit, the unit receive the abstract feature of the last layer master vector layer, form instantiation vector and are used for
Plant classification belonging to its feature of classified calculating.
The topography acquired in the step S1 is flower, leaf, fruit and the apparent plant organ image of feature of plant.
The apparent plant organ image of these features is acquired, the feature extraction being convenient for is convenient for identifying.
Taxonomic revision in the step S1 is that the plant data of acquisition are put into same file folder according to same type data
In.After congener data are put into a file, the characteristic of same plant can be obtained when carrying out machine learning
According to data will not be interlocked.
It is labeled as the file after the taxonomic revision in step S1 being numbered in the step S2, and in the form of text
Botanical name corresponding to each number is listed, the data set that can be used for training the vector neural network is formd.Institute
Stating the picture size in step S2 is 128*128.Multidimensional characteristic in the step S3 is to mention after several master vector units
It takes out, the feature with multiple dimensions.Vector algorithm in the step S3 routing algorithm between vector.The step S3
In substance parameter have blade texture, vein distribution, shape information, substance parameter by machine autonomous learning progress feature mention
It takes.
The above, only the preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto, it is any
Those skilled in the art within the technical scope disclosed by the invention, can without the variation that creative work is expected or
Replacement, should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be limited with claims
Subject to fixed protection scope.
Claims (8)
1. the multi-angle plants identification method based on vector neural network, including, it is characterised in that: the method includes walking as follows
It is rapid:
S1: collecting plant multi-angle image, passes through the quasi- multiple angles for identifying plant Various Seasonal of image acquisition device
Organ topography, collect and arrange image resource by plant classification carry out taxonomic revision it is spare;
S2: training set of images production, by the way that the plant image for collecting acquisition in step S1 is labeled as sub-category, and will
Image data after mark is uniformly adjusted to size of the same size;
S3: identification vector neural network model training, by being input to vector mind for treated in step S2 image data
In network, allowed the network to extract the multidimensional characteristic of plant image after successive ignition, and will be special by vector algorithm
Sign switchs to vector value, wherein vector field homoemorphism long a possibility that indicating its generic, and the direction of vector indicates that the classification of plant is real
Body parameter;
S4: by vector neural network that the training in step S3 obtains to identify floristics, belong to the figure for the plant to be identified
Picture independently extracts the feature with unchanged view angle through network, generates final feature vector, and according to this vector field homoemorphism
Length obtains identification classification results title to calculate Classification Loss after classified calculating;
S5: according to the classification results title calculated in step S4, searching for database or Internet resources data, and after showing search
Data details.
2. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
The topography acquired in step S1 is flower, leaf, fruit and the apparent plant organ image of feature of plant.
3. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
Taxonomic revision in step S1 is to be put into the plant data of acquisition in same file folder according to same type data.
4. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
It is labeled as the file after the taxonomic revision in step S1 being numbered in step S2, and lists each number in the form of text
Corresponding botanical name forms the data set that can be used for training the vector neural network.
5. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
Picture size in step S2 is 128*128.
6. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
Multidimensional characteristic in step S3 is to extract after several master vector units, the feature with multiple dimensions.
7. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
Vector algorithm in the step S3 routing algorithm between vector.
8. the multi-angle plants identification method according to claim 1 based on vector neural network, it is characterised in that: described
Substance parameter in step S3 has blade texture, vein distribution, shape information, and substance parameter is carried out special by machine autonomous learning
Sign is extracted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811263198.7A CN109446985B (en) | 2018-10-28 | 2018-10-28 | Multi-angle plant identification method based on vector neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811263198.7A CN109446985B (en) | 2018-10-28 | 2018-10-28 | Multi-angle plant identification method based on vector neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109446985A true CN109446985A (en) | 2019-03-08 |
CN109446985B CN109446985B (en) | 2021-06-04 |
Family
ID=65548963
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811263198.7A Active CN109446985B (en) | 2018-10-28 | 2018-10-28 | Multi-angle plant identification method based on vector neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109446985B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009034A (en) * | 2019-04-02 | 2019-07-12 | 中南大学 | Air-conditioner set type identifier method |
CN110321952A (en) * | 2019-07-02 | 2019-10-11 | 腾讯医疗健康(深圳)有限公司 | A kind of training method and relevant device of image classification model |
CN111898680A (en) * | 2020-07-31 | 2020-11-06 | 陈艳 | Biological identification method based on material inspection multi-view morphological image and deep learning |
CN112434579A (en) * | 2020-11-13 | 2021-03-02 | 深圳园林股份有限公司 | Method and device for extracting fallen leaf accommodating capacity of ground cover plant based on neural network |
CN113743576A (en) * | 2020-05-29 | 2021-12-03 | 精工爱普生株式会社 | Data acquisition method and apparatus, similarity calculation method and apparatus, and storage medium |
WO2022052052A1 (en) * | 2020-09-11 | 2022-03-17 | Siemens Aktiengesellschaft | Method and system for identifying objects |
JP7468155B2 (en) | 2020-05-29 | 2024-04-16 | セイコーエプソン株式会社 | Method, apparatus and computer program |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093247A (en) * | 2013-01-05 | 2013-05-08 | 山东华戎信息产业有限公司 | Automatic classification method for plant images |
CN105787503A (en) * | 2016-01-15 | 2016-07-20 | 四川德马克机器人科技有限公司 | Method for identifying plant leaf based on early state feature fusion |
CN103870816B (en) * | 2014-03-26 | 2016-11-23 | 中国科学院寒区旱区环境与工程研究所 | The method of the plants identification that a kind of discrimination is high |
CN106503687A (en) * | 2016-11-09 | 2017-03-15 | 合肥工业大学 | The monitor video system for identifying figures of fusion face multi-angle feature and its method |
CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
CN107239514A (en) * | 2017-05-19 | 2017-10-10 | 邓昌顺 | A kind of plants identification method and system based on convolutional neural networks |
CN107633038A (en) * | 2017-09-13 | 2018-01-26 | 浙江大学 | Tealeaves recognition methods and its system based on image recognition technology |
CN108038499A (en) * | 2017-12-06 | 2018-05-15 | 浙江大学 | A kind of seeds sorting technique and system based on deep learning |
CN108256568A (en) * | 2018-01-12 | 2018-07-06 | 宁夏智启连山科技有限公司 | A kind of plant species identification method and device |
-
2018
- 2018-10-28 CN CN201811263198.7A patent/CN109446985B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103093247A (en) * | 2013-01-05 | 2013-05-08 | 山东华戎信息产业有限公司 | Automatic classification method for plant images |
CN103870816B (en) * | 2014-03-26 | 2016-11-23 | 中国科学院寒区旱区环境与工程研究所 | The method of the plants identification that a kind of discrimination is high |
CN105787503A (en) * | 2016-01-15 | 2016-07-20 | 四川德马克机器人科技有限公司 | Method for identifying plant leaf based on early state feature fusion |
CN106503687A (en) * | 2016-11-09 | 2017-03-15 | 合肥工业大学 | The monitor video system for identifying figures of fusion face multi-angle feature and its method |
CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
CN107239514A (en) * | 2017-05-19 | 2017-10-10 | 邓昌顺 | A kind of plants identification method and system based on convolutional neural networks |
CN107633038A (en) * | 2017-09-13 | 2018-01-26 | 浙江大学 | Tealeaves recognition methods and its system based on image recognition technology |
CN108038499A (en) * | 2017-12-06 | 2018-05-15 | 浙江大学 | A kind of seeds sorting technique and system based on deep learning |
CN108256568A (en) * | 2018-01-12 | 2018-07-06 | 宁夏智启连山科技有限公司 | A kind of plant species identification method and device |
Non-Patent Citations (2)
Title |
---|
彭鸿元 等: "基于深度学习的植物识别技术的发展", 《电脑知识与技术》 * |
王路 等: "基于LVQ神经网络的植物种类识别", 《吉林大学学报(理学版)》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110009034A (en) * | 2019-04-02 | 2019-07-12 | 中南大学 | Air-conditioner set type identifier method |
CN110321952A (en) * | 2019-07-02 | 2019-10-11 | 腾讯医疗健康(深圳)有限公司 | A kind of training method and relevant device of image classification model |
CN110321952B (en) * | 2019-07-02 | 2024-02-09 | 腾讯医疗健康(深圳)有限公司 | Training method of image classification model and related equipment |
CN113743576A (en) * | 2020-05-29 | 2021-12-03 | 精工爱普生株式会社 | Data acquisition method and apparatus, similarity calculation method and apparatus, and storage medium |
CN113743575A (en) * | 2020-05-29 | 2021-12-03 | 精工爱普生株式会社 | Data acquisition method and apparatus, similarity calculation method and apparatus, and storage medium |
JP7468155B2 (en) | 2020-05-29 | 2024-04-16 | セイコーエプソン株式会社 | Method, apparatus and computer program |
CN113743576B (en) * | 2020-05-29 | 2024-05-07 | 精工爱普生株式会社 | Data acquisition method and device, similarity calculation method and device and storage medium |
CN113743575B (en) * | 2020-05-29 | 2024-05-07 | 精工爱普生株式会社 | Data acquisition method and device, similarity calculation method and device and storage medium |
CN111898680A (en) * | 2020-07-31 | 2020-11-06 | 陈艳 | Biological identification method based on material inspection multi-view morphological image and deep learning |
WO2022052052A1 (en) * | 2020-09-11 | 2022-03-17 | Siemens Aktiengesellschaft | Method and system for identifying objects |
CN112434579A (en) * | 2020-11-13 | 2021-03-02 | 深圳园林股份有限公司 | Method and device for extracting fallen leaf accommodating capacity of ground cover plant based on neural network |
Also Published As
Publication number | Publication date |
---|---|
CN109446985B (en) | 2021-06-04 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109446985A (en) | Multi-angle plants identification method based on vector neural network | |
Dong et al. | Photo quality assessment with DCNN that understands image well | |
Plummer et al. | Conditional image-text embedding networks | |
Kodirov et al. | Semantic autoencoder for zero-shot learning | |
CN112308158A (en) | Multi-source field self-adaptive model and method based on partial feature alignment | |
CN108108674A (en) | A kind of recognition methods again of the pedestrian based on joint point analysis | |
CN109543602A (en) | A kind of recognition methods again of the pedestrian based on multi-view image feature decomposition | |
CN112200211B (en) | Small sample fish identification method and system based on residual network and transfer learning | |
Yan et al. | A hybrid convolutional and recurrent deep neural network for breast cancer pathological image classification | |
CN109034953B (en) | Movie recommendation method | |
Li et al. | Multiple features with extreme learning machines for clothing image recognition | |
CN104966075B (en) | A kind of face identification method and system differentiating feature based on two dimension | |
CN108427740A (en) | A kind of Image emotional semantic classification and searching algorithm based on depth measure study | |
Qi et al. | Im2sketch: Sketch generation by unconflicted perceptual grouping | |
Zheng et al. | An efficient mobile model for insect image classification in the field pest management | |
Zhang et al. | Hybrid handcrafted and learned feature framework for human action recognition | |
Thanh-Nhan Nguyen et al. | A combination of deep learning and hand-designed feature for plant identification based on leaf and flower images | |
Hassan et al. | Vision based entomology: a survey | |
Ibrahim et al. | Improving Date Fruit Classification Using CycleGAN-Generated Dataset. | |
Martineau et al. | Effective training of convolutional neural networks for insect image recognition | |
Li et al. | Detection and identification of peach leaf diseases based on YOLO v5 improved model | |
Wang et al. | Deepbag: Recognizing handbag models | |
Hanafiah et al. | Flower recognition using deep convolutional neural networks | |
Deng et al. | Deep learning for gender recognition | |
Pandey et al. | Clustering of hierarchical image database to reduce inter-and intra-semantic gaps in visual space for finding specific image semantics |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |