CN110781902A - Robust binary attribute learning method and system - Google Patents
Robust binary attribute learning method and system Download PDFInfo
- Publication number
- CN110781902A CN110781902A CN201911028374.3A CN201911028374A CN110781902A CN 110781902 A CN110781902 A CN 110781902A CN 201911028374 A CN201911028374 A CN 201911028374A CN 110781902 A CN110781902 A CN 110781902A
- Authority
- CN
- China
- Prior art keywords
- binary code
- binary
- attribute
- discriminative
- registered
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 claims description 23
- 238000012360 testing method Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 12
- 230000002596 correlated effect Effects 0.000 claims description 9
- 230000001149 cognitive effect Effects 0.000 claims description 7
- 238000013527 convolutional neural network Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 6
- 230000009182 swimming Effects 0.000 description 4
- 230000000875 corresponding effect Effects 0.000 description 2
- 241000252229 Carassius auratus Species 0.000 description 1
- 241000282326 Felis catus Species 0.000 description 1
- 241001125046 Sardina pilchardus Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 235000019512 sardine Nutrition 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06V10/28—Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
-
- 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)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Electrically Operated Instructional Devices (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a robust binary attribute learning method and a robust binary attribute learning system, and belongs to the technical field of computer vision. The robust binary attribute learning method comprises the steps of taking each binary code as an attribute, representing physical characteristics of a target, constructing a binary attribute learner, constructing a related binary code characteristic learner according to a constructed binary attribute learning period, directly learning a plurality of related binary attributes, obtaining multi-dimensional binary code characteristics, fusing a plurality of binary code attributes and related multi-dimensional binary codes, inputting an image to be tested into the binary attribute learner and the related binary code characteristic learner, obtaining discriminative binary code characteristic representation of the image, comparing the discriminative binary code characteristic representation with a registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on similarity. The robust binary attribute learning method has better comprehensiveness and robustness, can hopefully improve the identification precision and efficiency, and has good popularization and application values.
Description
Technical Field
The invention relates to the technical field of computer vision, and particularly provides a robust binary attribute learning method and system.
Background
Binary codes have the advantages of easy storage, high calculation efficiency and the like. In the field of computer vision, an important application of binary codes is to represent attribute features as targets. However, the traditional binary code features have poor understandability and do not have good robustness. How to effectively solve the problems of poor understandability, poor robustness and the like of the existing binary code characteristics has important research significance and application value. Aiming at the problems of the existing binary code method, the invention provides an intuitive robust binary code learning method. The proposed binary code features have better comprehensibility and robustness, and are expected to improve the identification precision and efficiency.
Disclosure of Invention
The technical task of the invention is to provide a robust binary attribute learning method which has better comprehension and robustness and can hopefully improve the identification precision and efficiency aiming at the problems.
A further technical task of the present invention is to provide a robust binary attribute learning system.
In order to achieve the purpose, the invention provides the following technical scheme:
a robust binary attribute learning method includes the steps of enabling each binary code to serve as an attribute, representing physical characteristics of a target, constructing a binary attribute learning device, constructing a related binary code characteristic learning device according to a constructed binary attribute learning period, directly learning a plurality of related binary attributes, obtaining multi-dimensional binary code characteristics, fusing a plurality of binary code attributes and related multi-dimensional binary codes, inputting an image to be tested into the binary attribute learning device and the related binary code characteristic learning device, obtaining discriminative binary code characteristic representation of the image, comparing the discriminative binary code characteristic representation with a registered binary code template to obtain similarity of the two, and judging whether the discriminative binary code and the registered binary code are the same type or not based on the similarity.
The robust binary attribute learning method has better comprehensiveness and robustness, and can hopefully improve the identification precision and efficiency.
Preferably, the robust binary attribute learning method specifically includes the following steps:
s1, training phase
1) Constructing a binary attribute learning device;
2) constructing a relevant binary code learner;
and S2, in the testing stage, respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device to obtain the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
Preferably, in the process of constructing the binary attribute learner, the binary attributes of the target to be recognized are obtained based on the cognitive priori knowledge of the target in the sample, the sample is divided into a positive class and a negative class according to the value of each attribute, and training is carried out by using the densenet based on the collected sample and the attribute mark to obtain the binary attribute learner.
Preferably, the associated binary code learner is constructed by utilizing a convolutional neural network learning framework, a network framework adopts densenet, an associated loss function is introduced into a loss layer, as shown in formula (1), the associated binary code learner is obtained by minimizing an objective function,
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
Preferably, in the testing stage, the image to be tested is respectively input into the binary attribute learning device and the associated binary code learning device, the discriminative binary code feature representation of the image is obtained, the discriminative binary code feature representation is compared with the registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
A robust binary attribute learning system, the system comprising a training module and a testing module:
the training module is used for constructing a binary attribute learning device and an associated binary code learning device in a training stage;
the test module is used for respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device in the test stage, obtaining the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
Preferably, in the process of constructing the binary attribute learner, the training module acquires the binary attributes of the target to be recognized based on the cognitive priori knowledge of the target in the sample, divides the sample into a positive class and a negative class according to the value of each attribute, and trains the target by using the densenet based on the collected sample and the attribute mark to obtain the binary attribute learner.
Preferably, in the training module, a convolutional neural network learning framework is used for constructing a correlated binary code learner, a network architecture adopts densenet, a correlated loss function is introduced into a loss layer, as shown in formula (1), the correlated binary code learner is obtained by minimizing an objective function,
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
Preferably, the test module firstly inputs the image to be tested into the binary attribute learner and the associated binary code learner respectively in the test stage, discriminative binary code feature representation of the image is obtained, the discriminative binary code feature representation is compared with the registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
Compared with the prior art, the robust binary attribute learning method has the following outstanding beneficial effects: the robust binary attribute learning method has better comprehensibility and robustness, can hopefully improve the identification precision and efficiency, and has good popularization and application values.
Detailed Description
The robust binary attribute learning method and system of the present invention will be further described in detail with reference to the following embodiments.
Examples
The robust binary attribute learning method comprises the steps of taking each binary code as an attribute, representing physical characteristics of a target, constructing a binary attribute learner, constructing a related binary code characteristic learner according to a constructed binary attribute learning period, directly learning a plurality of related binary attributes, obtaining multi-dimensional binary code characteristics, fusing a plurality of binary code attributes and related multi-dimensional binary codes, inputting an image to be tested into the binary attribute learner and the related binary code characteristic learner, obtaining discriminative binary code characteristic representation of the image, comparing the discriminative binary code characteristic representation with a registered binary code template to obtain similarity of the discriminative binary code and the registered binary code, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
The robust binary attribute learning method specifically comprises the following steps:
s1, training phase
1) And constructing a binary attribute learner.
In the process of constructing the binary attribute learner, the binary attributes of the target to be recognized are obtained based on the cognitive priori knowledge of the target in the sample, the sample is divided into a positive class and a negative class according to the value of each attribute, and training is carried out by utilizing the densenet based on the collected sample and the attribute mark to obtain the binary attribute learner.
For example, in the present invention, assuming that the target to be identified includes a bird, a rain, and a cat, corresponding binary attributes may be designed, each of which may be considered an important component of the target. For example, in the above example, the cognitive priori information of the object by the person such as whether the object will fly, whether the object will swim, whether the object has a tail, and the like can be used as the binary attribute. For the swimming, a sample of a swimming target such as sardine or goldfish may be regarded as a positive class with an attribute value of 1, and a sample of a swimming target such as puppy, kitten or bird may be regarded as a negative class with an attribute value of 0. Training by using the densenet based on the collected samples and the attribute marks, wherein the training obtained densenet is the learner for judging whether the swimming attribute is expected or not. For the sample to be detected, the corresponding attribute value can be obtained only by inputting the image.
2) And constructing a relevant binary code learning device.
Constructing a correlation binary code learner by utilizing a convolutional neural network learning framework, introducing a correlation loss function into a loss layer by adopting a densenert in a network framework, obtaining the correlation binary code learner by minimizing an objective function as shown in a formula (1),
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
And S2, in the testing stage, respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device to obtain the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
In the testing stage, an image to be tested is firstly respectively input into the binary attribute learning device and the associated binary code learning device to obtain discriminative binary code feature representation of the image, the discriminative binary code feature representation is compared with a registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea name distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
A robust binary attribute learning system, the system comprising a training module and a testing module:
the training module is used for constructing a binary attribute learning device and an associated binary code learning device in a training stage;
the test module is used for respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device in the test stage, obtaining the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
Preferably, in the process of constructing the binary attribute learner, the training module acquires the binary attributes of the target to be recognized based on the cognitive priori knowledge of the target in the sample, divides the sample into a positive class and a negative class according to the value of each attribute, and trains the target by using the densenet based on the collected sample and the attribute mark to obtain the binary attribute learner.
Preferably, in the training module, a convolutional neural network learning framework is used for constructing a correlated binary code learner, a network architecture adopts densenet, a correlated loss function is introduced into a loss layer, as shown in formula (1), the correlated binary code learner is obtained by minimizing an objective function,
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
Preferably, the test module firstly inputs the image to be tested into the binary attribute learner and the associated binary code learner respectively in the test stage, discriminative binary code feature representation of the image is obtained, the discriminative binary code feature representation is compared with the registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
The above-described embodiments are merely preferred embodiments of the present invention, and general changes and substitutions by those skilled in the art within the technical scope of the present invention are included in the protection scope of the present invention.
Claims (9)
1. A robust binary attribute learning method is characterized in that: in the method, each binary code is used as an attribute to represent physical characteristics of a target, a binary attribute learning device is constructed, an associated binary code characteristic learning device is constructed according to a constructed binary attribute learning period, a plurality of relevant binary attributes are directly learned to obtain a multi-dimensional binary code characteristic, a plurality of binary code attributes and an associated multi-dimensional binary code are fused, an image to be tested is input into the binary attribute learning device and the associated binary code characteristic learning device to obtain a discriminative binary code characteristic representation of the image, the discriminative binary code characteristic representation is compared with a registered binary code template to obtain similarity of the two, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
2. The robust binary attribute learning method of claim 1, wherein: the method specifically comprises the following steps:
s1, training phase
1) Constructing a binary attribute learning device;
2) constructing a relevant binary code learner;
and S2, in the testing stage, respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device to obtain the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
3. The robust binary attribute learning method of claim 2, wherein: in the process of constructing the binary attribute learner, the binary attributes of the target to be recognized are obtained based on the cognitive priori knowledge of the target in the sample, the sample is divided into a positive class and a negative class according to the value of each attribute, and training is carried out by utilizing the densenet based on the collected sample and the attribute mark to obtain the binary attribute learner.
4. The robust binary attribute learning method of claim 3, wherein: constructing a correlation binary code learner by utilizing a convolutional neural network learning framework, introducing a correlation loss function into a loss layer by adopting a densenert in a network framework, obtaining the correlation binary code learner by minimizing an objective function as shown in a formula (1),
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
5. The robust binary attribute learning method of claim 4, wherein: in the testing stage, an image to be tested is firstly respectively input into the binary attribute learning device and the associated binary code learning device to obtain discriminative binary code feature representation of the image, the discriminative binary code feature representation is compared with a registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea name distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
6. A robust binary attribute learning system, characterized by: the system comprises a training module and a testing module:
the training module is used for constructing a binary attribute learning device and an associated binary code learning device in a training stage;
the test module is used for respectively inputting the image to be tested into the binary attribute learning device and the associated binary code learning device in the test stage, obtaining the discriminative binary code feature representation of the image, comparing the discriminative binary code feature representation with the registered binary code template to obtain the similarity of the discriminative binary code feature representation and the registered binary code template, and judging whether the discriminative binary code and the registered binary code are in the same class or not based on the similarity.
7. The robust binary attribute learning system of claim 6, wherein: in the process of constructing the binary attribute learner, the binary attributes of the target to be recognized are obtained based on the cognitive priori knowledge of the target in the sample, the sample is divided into a positive class and a negative class according to the value of each attribute, training is carried out by using a densenet based on the collected sample and the attribute marks, and the binary attribute learner is obtained.
8. The robust binary attribute learning system of claim 7, wherein: in the training module, a convolutional neural network learning framework is utilized to construct a correlated binary code learner, a network architecture adopts densenet, a correlated loss function is introduced into a loss layer, as shown in a formula (1), the correlated binary code learner is obtained by minimizing an objective function,
where t is the number of binary attributes, Y
(i，k)For the value of the kth attribute of all samples, X is the input image, is a matrix of nxd, W
kIs the associated weight vector, and is a hyperparameter.
9. The robust binary attribute learning system of claim 7, wherein: the test module firstly inputs an image to be tested into the binary attribute learning device and the associated binary code learning device respectively in a test stage, discriminative binary code feature representation of the image is obtained, the discriminative binary code feature representation is compared with a registered binary code template, the similarity between the discriminative binary code feature representation and the registered binary code template is obtained by sea distance measurement, and whether the discriminative binary code and the registered binary code are in the same class or not is judged based on the similarity.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911028374.3A CN110781902A (en) | 2019-10-28 | 2019-10-28 | Robust binary attribute learning method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911028374.3A CN110781902A (en) | 2019-10-28 | 2019-10-28 | Robust binary attribute learning method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110781902A true CN110781902A (en) | 2020-02-11 |
Family
ID=69386935
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911028374.3A Pending CN110781902A (en) | 2019-10-28 | 2019-10-28 | Robust binary attribute learning method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110781902A (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503106A (en) * | 2016-10-17 | 2017-03-15 | 北京工业大学 | A kind of image hash index construction method based on deep learning |
| CN106682233A (en) * | 2017-01-16 | 2017-05-17 | 华侨大学 | Method for Hash image retrieval based on deep learning and local feature fusion |
| CN108805157A (en) * | 2018-04-11 | 2018-11-13 | 南京理工大学 | Classifying Method in Remote Sensing Image based on the random supervision discrete type Hash in part |
-
2019
- 2019-10-28 CN CN201911028374.3A patent/CN110781902A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503106A (en) * | 2016-10-17 | 2017-03-15 | 北京工业大学 | A kind of image hash index construction method based on deep learning |
| CN106682233A (en) * | 2017-01-16 | 2017-05-17 | 华侨大学 | Method for Hash image retrieval based on deep learning and local feature fusion |
| CN108805157A (en) * | 2018-04-11 | 2018-11-13 | 南京理工大学 | Classifying Method in Remote Sensing Image based on the random supervision discrete type Hash in part |
Non-Patent Citations (1)
| Title |
|---|
| 高宪军: "半监督哈希算法研究", 《中国优秀博硕士学位论文全文数据库（硕士）信息科技辑》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110020651A (en) | Car plate detection localization method based on deep learning network | |
| CN110135295A (en) | A kind of unsupervised pedestrian recognition methods again based on transfer learning | |
| CN108897989A (en) | A kind of biological event abstracting method based on candidate events element attention mechanism | |
| CN110781663B (en) | Training method and device of text analysis model, text analysis method and device | |
| CN108960342B (en) | Image similarity calculation method based on improved Soft-Max loss function | |
| CN104765768A (en) | Mass face database rapid and accurate retrieval method | |
| CN104199840B (en) | Intelligent place name identification technology based on statistical model | |
| CN110175615B (en) | Model training method, domain-adaptive visual position identification method and device | |
| CN109583357A (en) | A kind of improvement LBP and the cascade face identification method of light weight convolutional neural networks | |
| CN107909053B (en) | Face detection method based on hierarchical learning cascade convolution neural network | |
| CN111369535B (en) | Cell detection method | |
| CN111079847A (en) | Remote sensing image automatic labeling method based on deep learning | |
| CN113704522A (en) | Artificial intelligence-based target image rapid retrieval method and system | |
| CN108446613A (en) | A kind of pedestrian's recognition methods again based on distance centerization and projection vector study | |
| CN116977633A (en) | Feature element segmentation model training method, feature element segmentation method and device | |
| CN111582373A (en) | Radiation source identification method based on weighted migration extreme learning machine algorithm | |
| CN115546553A (en) | Zero sample classification method based on dynamic feature extraction and attribute correction | |
| CN114329031A (en) | Fine-grained bird image retrieval method based on graph neural network and deep hash | |
| CN114579794A (en) | Multi-scale fusion landmark image retrieval method and system based on feature consistency suggestion | |
| CN111191027B (en) | Generalized zero sample identification method based on Gaussian mixture distribution (VAE) | |
| CN105701501A (en) | Trademark image identification method | |
| CN113792686A (en) | Vehicle weight identification method based on cross-sensor invariance of visual representation | |
| CN117516937A (en) | Rolling bearing unknown fault detection method based on multi-mode feature fusion enhancement | |
| CN110569764B (en) | Mobile phone model identification method based on convolutional neural network | |
| CN117197568A (en) | Zero sample image recognition method based on CLIP |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200211 |