CN115690566B - Deep sea animal new species identification method based on deep migration cluster learning - Google Patents
Deep sea animal new species identification method based on deep migration cluster learning Download PDFInfo
- Publication number
- CN115690566B CN115690566B CN202211299614.5A CN202211299614A CN115690566B CN 115690566 B CN115690566 B CN 115690566B CN 202211299614 A CN202211299614 A CN 202211299614A CN 115690566 B CN115690566 B CN 115690566B
- Authority
- CN
- China
- Prior art keywords
- classification network
- class classification
- deep
- network
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 241000894007 species Species 0.000 title claims abstract description 71
- 241001465754 Metazoa Species 0.000 title claims abstract description 62
- 230000005012 migration Effects 0.000 title claims abstract description 48
- 238000013508 migration Methods 0.000 title claims abstract description 48
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000007781 pre-processing Methods 0.000 claims abstract description 8
- 238000010276 construction Methods 0.000 claims abstract description 4
- 239000011159 matrix material Substances 0.000 claims description 49
- 239000013598 vector Substances 0.000 claims description 26
- 230000009467 reduction Effects 0.000 claims description 20
- 238000000513 principal component analysis Methods 0.000 claims description 16
- 238000003064 k means clustering Methods 0.000 claims description 7
- 230000001737 promoting effect Effects 0.000 claims description 3
- 230000002776 aggregation Effects 0.000 claims 2
- 238000010586 diagram Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 210000004556 brain Anatomy 0.000 description 1
- 230000009189 diving Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/80—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in fisheries management
- Y02A40/81—Aquaculture, e.g. of fish
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a deep sea animal new species identification method and system based on deep migration cluster learning, wherein the method comprises the following steps: shooting a deep sea animal photo and preprocessing to obtain a preprocessed picture; constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model; and carrying out species identification on the preprocessed picture based on the deep migration clustering learning model. The system comprises: shooting module, model construction module and recognition module. By using the method, the species classification marking can be carried out on the discovered deep sea animals, and the identification of new species of potential deep sea animals can be carried out. The deep sea animal new species identification method and system based on deep migration cluster learning can be widely applied to the field of image identification.
Description
Technical Field
The invention relates to the field of image recognition, in particular to a deep sea animal new species recognition method based on deep migration cluster learning.
Background
As the world countries continue to advance deep sea exploration, the world marine animal species has huge information base and is greatly increased with the day, and when unmanned remote control diving machines work in deep sea, it is quite difficult to purposefully discover and collect new species information of the deep sea animals in the deep sea in a short time. The traditional deep sea animal new species identification is more judged by an expert according to own knowledge reserve and network information retrieval, a great deal of manpower resources, time cost and subjective judgment depending on the expert are needed to be paid, if the expert fails to make effective judgment in time, the deep sea exploration research progress is slow or the deep sea animal new species discovery is missed, and finally, the deep sea animal new species identification is carried out by relying on manpower, so that the continuity of research work is required to be improved.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a deep sea animal new species identification method based on deep migration clustering learning, which can reduce the dependency of the discovery of new species of deep sea animals on experts, does not need to set any confidence threshold manually to give objective and effective judgment results, and can effectively learn new species characteristics in the process of completing the discovery of new species.
The first technical scheme adopted by the invention is as follows: a deep sea animal new species identification method based on deep migration cluster learning comprises the following steps:
shooting a deep sea animal photo and preprocessing to obtain a preprocessed picture;
constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
and carrying out species identification on the preprocessed picture based on the deep migration clustering learning model.
Further, the step of taking a photograph of the deep sea animal and performing pretreatment to obtain a pretreated picture specifically comprises the following steps:
taking a photograph of the deep sea animal based on the unmanned remote control submersible;
cutting the deep sea animal photo according to a preset size to obtain a cut picture;
and carrying out data enhancement on the cut picture to obtain a preprocessed picture.
Further, the step of constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration cluster learning model specifically comprises the following steps:
constructing a known class classification network based on the VGG16 and training to obtain a trained known class classification network;
constructing an unknown class classification network based on VGG16, training, and promoting the network to autonomously learn the characteristics of new species to obtain the trained unknown class classification network;
and integrating the known class classification network after training and the unknown class classification network after training to obtain the deep migration clustering learning model.
Further, the training steps of the known class classification network are as follows:
acquiring a training data set with category labels and inputting the training data set into a known category classification network;
outputting a first encoding vector matrix based on a feature extractor of the known class classification network;
performing dimension reduction on the first code vector matrix based on the principal component analysis dimension reduction layer to obtain a dimension reduced first code vector matrix;
clustering the first code vector matrix after dimension reduction based on a K-means clustering algorithm to obtain a first clustering center set;
performing single-heat coding on class labels in the training data set to obtain a target distribution matrix;
and aiming at minimizing the cross entropy loss of the probability distribution matrix and the target distribution matrix of the samples distributed to the first clustering center set, adjusting the parameters of the known class classification network (the network parameters of the feature extractor and the parameters of the PCA principal component analysis dimension reduction layer are adjusted by random gradient descent back propagation) to obtain the trained known class classification network.
Further, the expression of the probability distribution matrix that the samples are assigned to the first set of cluster centers is as follows:
P＝(1+||z a -U b || 2 ) -1
in the above formula, P represents that each sample is assigned to U b Probability distribution matrix of each cluster center, U b Representing a cluster center set, z a Representing the reduced dimension matrix of the code vectors.
Further, the training steps of the unknown class network are as follows:
constructing a target data set based on the preprocessed picture and inputting the target data set into an unknown class network;
extracting features of the target data set by using a feature extractor of the known class classification network to obtain a second coding vector matrix;
performing dimension reduction on the second coding vector matrix based on the principal component analysis dimension reduction layer to obtain a dimension reduced second coding vector matrix;
clustering the second code vector matrix after dimension reduction based on a K-means clustering algorithm to obtain a second cluster center set;
constructing a target probability distribution matrix based on a Bayesian criterion;
and taking the cross entropy loss of the probability distribution matrix and the target distribution matrix, which minimize the distribution of the samples to the second clustering center set, as a target, and adjusting the parameters of the unknown class classification network (the network parameters of the feature extractor and the parameters of the PCA principal component analysis dimension reduction layer are adjusted by random gradient descent back propagation) to obtain the trained unknown class classification network.
Further, the step of species identification on the preprocessed picture based on the deep migration clustering learning model specifically comprises the following steps:
selecting a target picture from the preprocessed pictures, and respectively inputting the target picture into a known class classification network and an unknown class classification network in the deep migration clustering learning model;
outputting a first classification label based on the unknown class classification network;
outputting a second category label based on the known category classification network;
searching for a deep sea animal photo corresponding to the second classification label in the training data set to obtain a search picture;
inputting the search pictures into an unknown class classification network to obtain search labels;
judging that the search tag is the same as the first classification tag, and recognizing that the deep sea animal in the target picture does not belong to a new species;
and judging that the search tag is different from the first classification tag, and identifying that the deep sea animal in the target picture belongs to a new species.
The second technical scheme adopted by the invention is as follows: a deep sea animal new species identification system based on deep migration cluster learning, comprising:
the shooting module is used for shooting a deep sea animal photo and preprocessing the photo to obtain a preprocessed picture;
the model construction module is used for constructing a known category classification network and an unknown category classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
and the identification module is used for identifying the species of the preprocessed picture based on the deep migration clustering learning model.
The method, the system, the device and the storage medium have the beneficial effects that: according to the invention, the deep sea animal photo is processed by acquiring the deep sea animal photo shot by the unmanned remote control submersible, and the species identification is performed by using the deep migration clustering learning model pre-trained by the deep sea animal species data set, so that the species classification marking of the found deep sea animal and the identification of the new species of the potential deep sea animal are realized in a short time, whether the species is a new species is judged by a technician in the field according to the threshold setting of the confidence level, the dependency of the discovery of the new species of the deep sea animal on an expert is reduced, and the characteristics of the found new species can be learned while the species are classified.
Drawings
FIG. 1 is a flow chart of steps of a method for identifying new species of deep sea animals based on deep migration cluster learning;
FIG. 2 is a schematic diagram of an application scenario of an embodiment of the present invention;
FIG. 3 is a schematic diagram of a VGG16 architecture with the classifier layer removed according to an embodiment of the invention;
FIG. 4 is a diagram of a training process of a known class classification network according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of an unknown class classification network training process in accordance with an embodiment of the present invention;
FIG. 6 is a schematic diagram of a general flow chart for species identification in accordance with an embodiment of the present invention.
Fig. 7 is a structural block diagram of a deep sea animal new species identification system based on deep migration cluster learning.
Detailed Description
The invention will now be described in further detail with reference to the drawings and to specific examples. The step numbers in the following embodiments are set for convenience of illustration only, and the order between the steps is not limited in any way, and the execution order of the steps in the embodiments may be adaptively adjusted according to the understanding of those skilled in the art.
As shown in fig. 1, the invention provides a deep sea animal new species identification method based on deep migration cluster learning, which comprises the following steps:
s1, shooting a deep sea animal photo and preprocessing to obtain a preprocessed picture;
s1.1, shooting a deep sea animal photo based on an unmanned remote control submersible;
in particular, referring to fig. 2, the unmanned submersible is controlled in real time by an operator of the offshore workstation, who is powered by electricityThe brain terminal remotely controlled unmanned submersible shoots and identifies i pictures of a target to form a target image set x i And transmitted to the offshore workstation server in real time.
S1.2, cutting the deep sea animal photo according to a preset size to obtain a cut picture;
specifically, the target image set x i The length and width are respectively 224 and 244.
S1.3, carrying out data enhancement on the cut picture to obtain a preprocessed picture.
Specifically, in ensuring the target image set x i For a target image set x with unchanged picture size i Randomly changing (e.g. turning over, stretching, cropping, deforming) the target image set x i J enhanced pictures can be obtained after each picture is subjected to data enhancement, and a target picture set x with the number of pictures being r (r=i+i×j) is formed r 。
S2, constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
s2.1, a known class classification network (VGG 16 network without classifier layer plus PCA dimension reduction layer and kmeans clustering layer form a classification network): VGG16 (visual geometry group network architecture 16, see figure three) with the classifier layer removed is used as a feature extractor for a known classification network omegaThe pretreatment picture length and width are 224 x 224, the sample number is a, the known class number is b, and the known class deep sea animal data set C a As input to VGG16, e.g. FishBase dataset, < >>The output code vector matrix M is reduced to b dimension by PCA (principal component analysis) to obtain a code vector matrix z a By the method of z a Clustering by K-means (K-means clustering algorithm) to obtain a cluster center set U b . Each sample is assigned to U b The probability distribution matrix P for each cluster center is defined as follows:
P＝(1+||z a -U b || 2 ) -1
c is C a And (3) carrying out one-hot coding (single-hot coding) on the label of the model to obtain a target distribution matrix q, minimizing the cross entropy loss of P and q, and adjusting the parameters of the model by random gradient descent back propagation, wherein the known classification network omega pre-training flow is shown in a figure 4.
S2.2, unknown class classification network: usingFor x r Feature extraction is carried out to obtain a coding vector matrix N r For N r Reducing to b+1 dimensions using PCA (principal component analysis) to obtain a matrix Y of encoded vectors r By means of Y r Clustering by K-means (K-means clustering algorithm) to obtain a cluster center set O b+1 Constructing a probability distribution matrix p', with individual samples assigned to O b+1 The probability p' (r|k) of each cluster center is defined as follows:
p′ (r|k) ＝(1+||Y r -O b+1 || 2 ) -1
constructing a target probability distribution matrix q' based on Bayesian criteria, with individual samples assigned to O b+1 The target probability q' (r|k) for each cluster center is defined as follows:
the training flow of the unknown class learning classification network ρ is shown in fig. 5.
And S3, performing species identification on the preprocessed picture based on the deep migration clustering learning model.
Specifically, after ρ learning is completed, x is the sum of i Selecting one target picture x as the input of rho to obtain a classification label Y, then taking x as the input of omega to obtain the classification label Y, searching the deep sea species picture F corresponding to the label Y in the training data set, inputting F into rho to obtain the label Y', and if the target picture is markedIf the label Y' is the same as the label Y, the current target is not considered to be a new species, and the label Y is marked on the current target; if the tags Y' and Y are not identical, it is stated that the presence of a decisive feature between the current target and the animal species most likely to be categorized results in a distinction between the two, considering the deep-sea animal as belonging to a new species. Finally, species identification is performed by using a deep migration clustering learning model pre-trained by a deep sea animal species data set, as shown in fig. 6.
As shown in fig. 7, a deep sea animal new species identification system based on deep migration cluster learning includes:
the shooting module is used for shooting a deep sea animal photo and preprocessing the photo to obtain a preprocessed picture;
the model construction module is used for constructing a known category classification network and an unknown category classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
and the identification module is used for identifying the species of the preprocessed picture based on the deep migration clustering learning model.
The content in the method embodiment is applicable to the system embodiment, the functions specifically realized by the system embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
Deep sea animal new species identification device based on deep migration cluster learning:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a deep sea animal new species identification method based on deep migration cluster learning as described above.
The content in the method embodiment is applicable to the embodiment of the device, and the functions specifically realized by the embodiment of the device are the same as those of the method embodiment, and the obtained beneficial effects are the same as those of the method embodiment.
A storage medium having stored therein instructions executable by a processor, characterized by: the processor-executable instructions, when executed by the processor, are for implementing a deep sea animal new species identification method based on deep migration cluster learning as described above.
The content in the method embodiment is applicable to the storage medium embodiment, and functions specifically implemented by the storage medium embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
While the preferred embodiment of the present invention has been described in detail, the invention is not limited to the embodiment, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the invention, and these modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.
Claims (5)
1. A deep sea animal new species identification method based on deep migration cluster learning is characterized by comprising the following steps:
shooting a deep sea animal photo and preprocessing to obtain a preprocessed picture;
constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
species identification is carried out on the preprocessed pictures based on the deep migration clustering learning model;
the step of constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model specifically comprises the following steps:
constructing a known class classification network based on the VGG16 and training to obtain a trained known class classification network;
constructing an unknown class classification network based on VGG16, training, and promoting the network to autonomously learn the characteristics of new species to obtain the trained unknown class classification network;
integrating the known class classification network after training and the unknown class classification network after training to obtain a deep migration clustering learning model;
the training steps of the unknown class classification network are as follows:
constructing a target data set based on the preprocessed picture and inputting the target data set into an unknown class network;
extracting features of the target data set by using a feature extractor of the known class classification network to obtain a second coding vector matrix;
performing dimension reduction on the second coding vector matrix based on the principal component analysis dimension reduction layer to obtain a dimension reduced second coding vector matrix;
clustering the second code vector matrix after dimension reduction based on a K-means clustering algorithm to obtain a second cluster center set;
constructing a target probability distribution matrix based on a Bayesian criterion;
the method comprises the steps of taking a probability distribution matrix and a target distribution matrix, which minimize samples to be distributed to a second aggregation center set, as targets, and adjusting parameters of an unknown class classification network through random gradient descent back propagation to obtain a trained unknown class classification network;
parameters of the unknown class classification network comprise network parameters of a feature extractor, parameters of a PCA principal component analysis dimension reduction layer and a kmeans cluster center set;
the step of species identification on the preprocessed picture based on the deep migration clustering learning model specifically comprises the following steps:
selecting a target picture from the preprocessed pictures, and respectively inputting the target picture into a known class classification network and an unknown class classification network in the deep migration clustering learning model;
outputting a first classification label based on the unknown class classification network;
outputting a second category label based on the known category classification network;
searching for a deep sea animal photo corresponding to the second classification label in the training data set to obtain a search picture;
inputting the search pictures into an unknown class classification network to obtain search labels;
judging that the search tag is the same as the first classification tag, and recognizing that the deep sea animal in the target picture does not belong to a new species;
and judging that the search tag is different from the first classification tag, and identifying that the deep sea animal in the target picture belongs to a new species.
2. The method for identifying new species of deep sea animals based on deep migration cluster learning of claim 1, wherein the steps of taking photographs of the deep sea animals and preprocessing the photographs to obtain preprocessed photographs comprise:
taking a photograph of the deep sea animal based on the unmanned remote control submersible;
cutting the deep sea animal photo according to a preset size to obtain a cut picture;
and carrying out data enhancement on the cut picture to obtain a preprocessed picture.
3. The deep sea animal new species identification method based on deep migration cluster learning of claim 2, wherein the training step of the known class classification network is as follows:
acquiring a training data set with category labels and inputting the training data set into a known category classification network;
outputting a first encoding vector matrix based on a feature extractor of the known class classification network;
performing dimension reduction on the first code vector matrix based on the principal component analysis dimension reduction layer to obtain a dimension reduced first code vector matrix;
clustering the first code vector matrix after dimension reduction based on a K-means clustering algorithm to obtain a first clustering center set;
performing single-heat coding on class labels in the training data set to obtain a target distribution matrix;
and aiming at minimizing the cross entropy loss of the probability distribution matrix and the target distribution matrix of the samples distributed to the first clustering center set, and adjusting the parameters of the known class classification network through random gradient descent back propagation to obtain the trained known class classification network.
4. A deep sea animal new species identification method based on deep migration cluster learning as claimed in claim 3, wherein the expression of the probability distribution matrix of the samples assigned to the first cluster center set is as follows:
P＝(t+||z a -U b || 2 ) -1
in the above formula, P represents that each sample is assigned to U b Probability distribution matrix of each cluster center, U b Representing a cluster center set, z a Representing the reduced dimension matrix of the code vectors.
5. Deep sea animal new species identification system based on deep migration cluster learning, which is characterized by comprising:
the shooting module is used for shooting a deep sea animal photo and preprocessing the photo to obtain a preprocessed picture;
the model construction module is used for constructing a known category classification network and an unknown category classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model;
the identification module is used for identifying the species of the preprocessed picture based on the deep migration clustering learning model;
the step of constructing a known class classification network and an unknown class classification network, training the network and learning new species characteristics to obtain a deep migration clustering learning model specifically comprises the following steps:
constructing a known class classification network based on the VGG16 and training to obtain a trained known class classification network;
constructing an unknown class classification network based on VGG16, training, and promoting the network to autonomously learn the characteristics of new species to obtain the trained unknown class classification network;
integrating the known class classification network after training and the unknown class classification network after training to obtain a deep migration clustering learning model;
the training steps of the unknown class classification network are as follows:
constructing a target data set based on the preprocessed picture and inputting the target data set into an unknown class network;
extracting features of the target data set by using a feature extractor of the known class classification network to obtain a second coding vector matrix;
performing dimension reduction on the second coding vector matrix based on the principal component analysis dimension reduction layer to obtain a dimension reduced second coding vector matrix;
clustering the second code vector matrix after dimension reduction based on a K-means clustering algorithm to obtain a second cluster center set;
constructing a target probability distribution matrix based on a Bayesian criterion;
the method comprises the steps of taking a probability distribution matrix and a target distribution matrix, which minimize samples to be distributed to a second aggregation center set, as targets, and adjusting parameters of an unknown class classification network through random gradient descent back propagation to obtain a trained unknown class classification network;
parameters of the unknown class classification network comprise network parameters of a feature extractor, parameters of a PCA principal component analysis dimension reduction layer and a kmeans cluster center set;
the step of species identification on the preprocessed picture based on the deep migration clustering learning model specifically comprises the following steps:
selecting a target picture from the preprocessed pictures, and respectively inputting the target picture into a known class classification network and an unknown class classification network in the deep migration clustering learning model;
outputting a first classification label based on the unknown class classification network;
outputting a second category label based on the known category classification network;
searching for a deep sea animal photo corresponding to the second classification label in the training data set to obtain a search picture;
inputting the search pictures into an unknown class classification network to obtain search labels;
judging that the search tag is the same as the first classification tag, and recognizing that the deep sea animal in the target picture does not belong to a new species;
and judging that the search tag is different from the first classification tag, and identifying that the deep sea animal in the target picture belongs to a new species.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211299614.5A CN115690566B (en) | 2022-10-24 | 2022-10-24 | Deep sea animal new species identification method based on deep migration cluster learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211299614.5A CN115690566B (en) | 2022-10-24 | 2022-10-24 | Deep sea animal new species identification method based on deep migration cluster learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115690566A CN115690566A (en) | 2023-02-03 |
| CN115690566B true CN115690566B (en) | 2024-01-19 |
Family
ID=85066652
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211299614.5A Active CN115690566B (en) | 2022-10-24 | 2022-10-24 | Deep sea animal new species identification method based on deep migration cluster learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115690566B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116468984B (en) * | 2023-03-10 | 2023-10-27 | 衡阳师范学院 | Construction method of movable object detection model, detection model and detection method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110580496A (en) * | 2019-07-11 | 2019-12-17 | 南京邮电大学 | Deep migration learning system and method based on entropy minimization |
| WO2021022816A1 (en) * | 2019-08-07 | 2021-02-11 | 南京硅基智能科技有限公司 | Intent identification method based on deep learning network |
| CN113283514A (en) * | 2021-05-31 | 2021-08-20 | 高新兴科技集团股份有限公司 | Unknown class classification method, device and medium based on deep learning |
| CN113326289A (en) * | 2021-08-02 | 2021-08-31 | 山东大学 | Rapid cross-modal retrieval method and system for incremental data carrying new categories |
| CN113705580A (en) * | 2021-08-31 | 2021-11-26 | 西安电子科技大学 | Hyperspectral image classification method based on deep migration learning |
| CN113705507A (en) * | 2021-09-02 | 2021-11-26 | 上海交通大学 | Mixed reality open set human body posture recognition method based on deep learning |
-
2022
- 2022-10-24 CN CN202211299614.5A patent/CN115690566B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110580496A (en) * | 2019-07-11 | 2019-12-17 | 南京邮电大学 | Deep migration learning system and method based on entropy minimization |
| WO2021022816A1 (en) * | 2019-08-07 | 2021-02-11 | 南京硅基智能科技有限公司 | Intent identification method based on deep learning network |
| CN113283514A (en) * | 2021-05-31 | 2021-08-20 | 高新兴科技集团股份有限公司 | Unknown class classification method, device and medium based on deep learning |
| CN113326289A (en) * | 2021-08-02 | 2021-08-31 | 山东大学 | Rapid cross-modal retrieval method and system for incremental data carrying new categories |
| CN113705580A (en) * | 2021-08-31 | 2021-11-26 | 西安电子科技大学 | Hyperspectral image classification method based on deep migration learning |
| CN113705507A (en) * | 2021-09-02 | 2021-11-26 | 上海交通大学 | Mixed reality open set human body posture recognition method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| Deep Convolutional Transform Learning - Extended version;Maggu, J et.al;《arXiv》;第1-10页 * |
| Learning to Discover Novel Visual Categories via Deep Transfer Clustering;Kai Han et.al;《2019 IEEE/CVF International Conference on Computer Vision (ICCV)》;第8400-8408页 * |
| 基于CNN迁移学习的示功图图形分类预警;朱顺德;徐增兴;刘一璠;;中国管理信息化(09);第155-157页 * |
| 基于深度学习的属性网络表示学习方法研究;王胜;《中国优秀硕士学位论文全文数据库 基础科学辑》;第A002-36页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115690566A (en) | 2023-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106980641B (en) | Unsupervised Hash quick picture retrieval system and unsupervised Hash quick picture retrieval method based on convolutional neural network | |
| CN110909820A (en) | Image classification method and system based on self-supervision learning | |
| CN115690566B (en) | Deep sea animal new species identification method based on deep migration cluster learning | |
| CN114092742B (en) | Multi-angle-based small sample image classification device and method | |
| CN114677687A (en) | ViT and convolutional neural network fused writing brush font type rapid identification method | |
| CN111476289A (en) | Fish shoal identification method, device, equipment and storage medium based on feature library | |
| CN116206334A (en) | Wild animal identification method and device | |
| CN114742224A (en) | Pedestrian re-identification method and device, computer equipment and storage medium | |
| CN114782997A (en) | Pedestrian re-identification method and system based on multi-loss attention adaptive network | |
| CN111242114B (en) | Character recognition method and device | |
| CN116108217B (en) | Fee evasion vehicle similar picture retrieval method based on depth hash coding and multitask prediction | |
| CN116935411A (en) | Radical-level ancient character recognition method based on character decomposition and reconstruction | |
| Song et al. | Text Siamese network for video textual keyframe detection | |
| CN113192108B (en) | Man-in-loop training method and related device for vision tracking model | |
| CN112905832B (en) | Complex background fine-grained image retrieval system and method | |
| CN115599953A (en) | Training method and retrieval method of video text retrieval model and related equipment | |
| CN111553202B (en) | Training method, detection method and device for neural network for living body detection | |
| CN114581769A (en) | Method for identifying houses under construction based on unsupervised clustering | |
| CN113553947A (en) | Method and device for re-identifying pedestrian by embedding multi-mode into generation description and electronic equipment | |
| CN113822304B (en) | Small sample learning method based on deep learning | |
| CN116843997B (en) | Model training method, cell image labeling method, device, equipment and storage medium | |
| CN116842479B (en) | Image processing method, device, computer equipment and storage medium | |
| CN111461207B (en) | Picture identification model training system and method | |
| CN114842233A (en) | Sequence random network image classification method | |
| Chen et al. | Compare Network with Task-dependent Embeddings for Few-shot Learning |
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 |