CN112200598A - Picture advertisement identification method and device and computer equipment - Google Patents
Picture advertisement identification method and device and computer equipment Download PDFInfo
- Publication number
- CN112200598A CN112200598A CN202010932388.4A CN202010932388A CN112200598A CN 112200598 A CN112200598 A CN 112200598A CN 202010932388 A CN202010932388 A CN 202010932388A CN 112200598 A CN112200598 A CN 112200598A
- Authority
- CN
- China
- Prior art keywords
- picture
- picture advertisement
- advertisement
- embedding characteristics
- mode
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0248—Avoiding fraud
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Finance (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- Evolutionary Computation (AREA)
- Accounting & Taxation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Entrepreneurship & Innovation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Game Theory and Decision Science (AREA)
- Molecular Biology (AREA)
- Economics (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a picture advertisement identification method, a picture advertisement identification device and computer equipment. Wherein the method comprises the following steps: the method comprises the steps of obtaining word embedding characteristics in the picture advertisement, extracting picture embedding characteristics in the picture advertisement, constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics, and identifying whether the picture advertisement is the cheating picture advertisement or not according to the constructed two-classification model of the cheating picture advertisement. Through the mode, the accuracy rate of identifying the cheating picture advertisements can be improved.
Description
Technical Field
The invention relates to the technical field of picture advertisement identification, in particular to a picture advertisement identification method, a picture advertisement identification device and computer equipment.
Background
The picture advertisement is an advertisement for transmitting advertisement information in a form of graph-based representation. The picture advertisement is generally called flag advertisement, which is composed of figures and characters and has strong image.
The picture advertisement is characterized in that the intention to be expressed is framed at the static moment of a certain scene, and the picture advertisement has strong readability and high visibility due to the intuitiveness of the shape, the authenticity of the color and the contrast of the light and shadow, and is easy to obtain the identity of a user. The pictures are superior to the characters in the expression of the form beauty and the form beauty which are specifically sensible to the objective outside. Therefore, the picture advertisement has great propagation advantages in having such a large advantage of the picture.
However, in the existing picture advertisement Recognition scheme, when a fraudulent picture advertisement is recognized, generally, a picture OCR (Optical Character Recognition) result is implemented by text semantic detection and LOGO detection in a picture, but the picture Optical Character Recognition result is implemented by text semantic detection and LOGO detection in the picture, and since environmental scene information of the fraudulent picture advertisement is not combined, a situation that the fraudulent picture advertisement is misjudged frequently occurs, and the accuracy of recognizing the fraudulent picture advertisement is general.
Disclosure of Invention
In view of this, the present invention provides a picture advertisement recognition method, device and computer device, which can improve the accuracy of recognizing fraudulent picture advertisements.
According to an aspect of the present invention, there is provided a picture advertisement recognition method, including: acquiring word embedding characteristics in the picture advertisement; extracting graph embedding features in the picture advertisement; constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics; and identifying whether the picture advertisement is a fraudulent picture advertisement or not according to the established binary classification model of the fraudulent picture advertisement.
Wherein, the obtaining of word embedding characteristics in the picture advertisement includes: the method comprises the steps of obtaining characters in the picture advertisement in a character recognition mode, and obtaining word embedding characteristics in the picture advertisement in a general semantic representation mode.
Wherein, the extracting of the graph embedding features in the picture advertisement comprises: extracting environmental scene information from the picture advertisement according to a preset environmental scene model, and extracting the picture embedding characteristics in the picture advertisement by adopting a convolutional neural network mode based on the environmental scene information.
The establishing of the two classification models of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics comprises the following steps: and carrying out differential operation and normalization operation on the word embedding characteristics and the graph embedding characteristics, and constructing a binary classification model of the cheating picture advertisement by taking the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation as a training input mode of a multi-mode deep semantic interaction neural network.
After the identifying whether the picture advertisement is the fraud picture advertisement according to the established binary classification model of the fraud picture advertisement, the method further comprises the following steps: and training and updating the established binary model of the cheating picture advertisement through preset times of iteration.
According to another aspect of the present invention, there is provided a picture advertisement recognition apparatus, including: the device comprises an acquisition module, an extraction module, a construction module and an identification module; the acquisition module is used for acquiring word embedding characteristics in the picture advertisement; the extraction module is used for extracting the graph embedding characteristics in the picture advertisement; the building module is used for building a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics; and the identification module is used for identifying whether the picture advertisement is a fraudulent picture advertisement according to the established binary classification model of the fraudulent picture advertisement.
The obtaining module is specifically configured to: the method comprises the steps of obtaining characters in the picture advertisement in a character recognition mode, and obtaining word embedding characteristics in the picture advertisement in a general semantic representation mode.
Wherein, the extraction module is specifically configured to: extracting environmental scene information from the picture advertisement according to a preset environmental scene model, and extracting the picture embedding characteristics in the picture advertisement by adopting a convolutional neural network mode based on the environmental scene information.
Wherein the building block is specifically configured to: and carrying out differential operation and normalization operation on the word embedding characteristics and the graph embedding characteristics, and constructing a binary classification model of the cheating picture advertisement by taking the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation as a training input mode of a multi-mode deep semantic interaction neural network.
Wherein, the picture advertisement recognition device further comprises: an update module; and the updating module is used for training and updating the established binary model of the cheating picture advertisement through iteration of preset times.
According to yet another aspect of the present invention, there is provided a computer apparatus comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the picture advertisement recognition method according to any one of the above.
According to a further aspect of the present invention, there is provided a computer readable storage medium storing a computer program which, when executed by a processor, implements a picture advertisement recognition method as described in any one of the above.
The scheme can obtain the word embedding characteristics in the picture advertisement, extract the picture embedding characteristics in the picture advertisement, construct a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics, identify whether the picture advertisement is the cheating picture advertisement according to the constructed two-classification model of the cheating picture advertisement, and improve the accuracy of identifying the cheating picture advertisement.
Furthermore, the above scheme can adopt a character recognition mode to obtain characters in the picture advertisement, and obtain the word embedding characteristics in the picture advertisement through a general semantic representation model mode, so that the advantage that the general semantic representation model can predict the forward and backward context of words, and the accuracy rate of the obtained word embedding characteristics can be improved.
Furthermore, the above scheme can extract the environmental scene information from the picture advertisement according to the preset environmental scene model, and extract the picture embedding feature in the picture advertisement by adopting the convolution neural network mode based on the environmental scene information, so that the advantage that the accuracy of the extracted picture embedding feature can be improved can be realized due to the fact that the environmental scene information in the picture advertisement and the information of the picture advertisement context can be reserved by the convolution neural network.
Furthermore, according to the scheme, the word embedding characteristics and the graph embedding characteristics can be subjected to differential operation and normalization operation, and the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation are used as a training input mode of the multi-mode deep semantic interaction neural network to construct the two classification models of the cheating picture advertisement.
Furthermore, the scheme can train and update the established binary model of the cheating picture advertisement through preset times of iteration, and the method has the advantage that the accuracy rate of identifying the cheating picture advertisement can be further improved.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a flow chart illustrating an embodiment of a method for identifying a picture advertisement according to the present invention;
FIG. 2 is a flow chart illustrating an alternative embodiment of a method for identifying a picture advertisement of the present invention;
FIG. 3 is a schematic structural diagram of an embodiment of a picture advertisement recognition apparatus according to the present invention;
FIG. 4 is a schematic structural diagram of an image advertisement recognition device according to another embodiment of the present invention;
FIG. 5 is a schematic structural diagram of an embodiment of the computer apparatus of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. It is to be noted that the following examples are only illustrative of the present invention, and do not limit the scope of the present invention. Similarly, the following examples are only some but not all examples of the present invention, and all other examples obtained by those skilled in the art without any inventive work are within the scope of the present invention.
The invention provides a picture advertisement identification method, which can improve the accuracy of identifying fraudulent picture advertisements.
Referring to fig. 1, fig. 1 is a schematic flow chart of an embodiment of a picture advertisement recognition method according to the present invention. It should be noted that the method of the present invention is not limited to the flow sequence shown in fig. 1 if the results are substantially the same. As shown in fig. 1, the method comprises the steps of:
s101: acquiring a word embedding (embedding) feature in the picture advertisement.
The obtaining of the word embedding feature in the picture advertisement may include:
the method has the advantages that the characters in the picture advertisement are obtained by adopting a character recognition mode, and the word embedding characteristics in the picture advertisement are obtained by a Bert (general semantic Representation model) mode, so that the method has the advantage that the general semantic Representation model can predict the forward and backward contexts of words, and the accuracy of the obtained word embedding characteristics can be improved.
In this embodiment, the character recognition method may be an optical character recognition method, or other recognition methods, and the invention is not limited thereto.
S102: and extracting graph embedding characteristics in the picture advertisement.
Wherein, the extracting of the graph embedding feature in the graph advertisement may include:
the method has the advantages that the environmental scene information in the picture advertisement and the information of the context of the picture advertisement can be combined, and the accuracy of the extracted picture embedding feature can be improved.
In this embodiment, the picture corresponding to the picture advertisement may be divided into a plurality of picture blocks, picture features may be extracted from each picture block, a semantic context relationship between each picture block is determined according to the picture features, the picture features are connected together according to the semantic context relationship in a sequential manner expressed by a sentence semantic context, and environmental scene information may be extracted from the picture advertisement according to the semantic context relationship of the connected picture features and a preset environmental scene model, or environmental scene information may be extracted from the picture advertisement in other manners, which is not limited in the present invention.
In this embodiment, the plurality of picture blocks may be picture blocks of the same size, picture blocks of different sizes, picture blocks of the same size in part, and the like, and the present invention is not limited thereto.
In this embodiment, the preset environment scene model may include environment scene models of scenic spots, libraries, offices, homes, relatives, people, rooms, docks, seasides, learning, class, dining, banks, prisons, factories, hospitals, supermarkets, rural areas, markets, rivers, schools, cities, hotels, communities, parks, airports, stations, tourism, gymnasiums, beauty treatment, and the like, which is not limited in the present invention.
S103: and constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the graph embedding characteristics.
The constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics can comprise:
the word embedding characteristics and the graph embedding characteristics are subjected to differential operation and normalization operation, and the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation are used as a training input mode of a multi-mode deep semantic interaction neural network to construct a two-classification model of the cheating picture advertisement.
S104: and identifying whether the picture advertisement is a fraudulent picture advertisement or not according to the established binary classification model of the fraudulent picture advertisement.
After identifying whether the picture advertisement is a fraudulent picture advertisement according to the established binary classification model of the fraudulent picture advertisement, the method may further include:
the established binary model of the cheating picture advertisement is trained and updated through iteration of preset times, so that the method has the advantage that the accuracy rate of identifying the cheating picture advertisement can be further improved.
It can be found that, in this embodiment, the word embedding feature in the picture advertisement can be obtained, the graph embedding feature in the picture advertisement can be extracted, a multi-mode deep semantic interaction neural network mode can be adopted based on the word embedding feature and the graph embedding feature to construct a two-classification model of the cheating picture advertisement, and whether the picture advertisement is the cheating picture advertisement can be identified according to the constructed two-classification model of the cheating picture advertisement, so that the accuracy of identifying the cheating picture advertisement can be improved.
Furthermore, in this embodiment, a character recognition mode may be adopted to obtain characters in the picture advertisement, and a general semantic representation model mode is adopted to obtain word embedding features in the picture advertisement, which has the advantage that since the general semantic representation model can predict forward and backward contexts of words, the accuracy of the obtained word embedding features can be improved.
Further, in this embodiment, the environmental scene information may be extracted from the picture advertisement according to a preset environmental scene model, and the graph embedding feature in the picture advertisement may be extracted in a convolutional neural network manner based on the environmental scene information, which is advantageous in that the accuracy of the extracted graph embedding feature may be improved because the environmental scene information in the picture advertisement and the information of the context of the picture advertisement may be retained by the convolutional neural network.
Further, in this embodiment, the word embedding feature and the graph embedding feature may be subjected to a difference operation and a normalization operation, and the word embedding feature and the graph embedding feature subjected to the difference operation and the normalization operation may be used as a training input of the multi-modal deep semantic interaction neural network to construct a two-classification model of the fraudulent picture advertisement.
Referring to fig. 2, fig. 2 is a schematic flow chart of another embodiment of the picture advertisement recognition method according to the present invention. In this embodiment, the method includes the steps of:
s201: and acquiring word embedding characteristics in the picture advertisement.
As described above in S101, further description is omitted here.
S202: and extracting graph embedding characteristics in the picture advertisement.
As described above in S102, further description is omitted here.
S203: and constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the graph embedding characteristics.
As described above in S103, which is not described herein.
S204: and identifying whether the picture advertisement is a fraudulent picture advertisement or not according to the established binary classification model of the fraudulent picture advertisement.
As described above in S104, and will not be described herein.
S205: and training and updating the established binary classification model of the cheating picture advertisement through preset times of iteration.
It can be found that, in this embodiment, the established binary model of the fraudulent picture advertisement can be trained and updated through a preset number of iterations, which has the advantage of further improving the accuracy of identifying the fraudulent picture advertisement.
The invention also provides a picture advertisement recognition device, which can improve the accuracy of recognizing the cheating picture advertisement.
Referring to fig. 3, fig. 3 is a schematic structural diagram of an embodiment of a picture advertisement recognition device according to the present invention. In this embodiment, the picture advertisement recognition device 30 includes an obtaining module 31, an extracting module 32, a constructing module 33, and a recognition module 34.
The obtaining module 31 is configured to obtain word embedding characteristics in the picture advertisement.
The extracting module 32 is configured to extract a graph embedding feature in the picture advertisement.
The building module 33 is configured to build a two-classification model of the cheating picture advertisement by using a neural network manner of multi-modal deep semantic interaction based on the word embedding feature and the picture embedding feature.
The identifying module 34 is configured to identify whether the picture advertisement is a fraudulent picture advertisement according to the established binary model of the fraudulent picture advertisement.
Optionally, the obtaining module 31 may be specifically configured to:
the method comprises the steps of obtaining characters in the picture advertisement in a character recognition mode, and obtaining word embedding characteristics in the picture advertisement in a general semantic representation mode.
Optionally, the extracting module 32 may be specifically configured to:
extracting environmental scene information from the picture advertisement according to a preset environmental scene model, and extracting the picture embedding characteristics in the picture advertisement by adopting a convolutional neural network mode based on the environmental scene information.
Optionally, the building block 33 may be specifically configured to:
and carrying out differential operation and normalization operation on the word embedding characteristics and the graph embedding characteristics, and constructing a binary classification model of the cheating picture advertisement by taking the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation as a training input mode of a multi-mode deep semantic interaction neural network.
Referring to fig. 4, fig. 4 is a schematic structural diagram of an image advertisement recognition device according to another embodiment of the present invention. Different from the previous embodiment, the picture advertisement recognition device 40 of the present embodiment further includes an updating module 41.
The updating module 41 is configured to train and update the constructed binary model of the fraudulent picture advertisement through a preset number of iterations.
Each unit module of the picture advertisement recognition device 30/40 can respectively execute the corresponding steps in the above method embodiments, and therefore, the detailed description of each unit module is omitted here, and please refer to the description of the corresponding steps above.
The present invention further provides a computer device, as shown in fig. 5, comprising: at least one processor 51; and a memory 52 communicatively coupled to the at least one processor 51; the memory 52 stores instructions executable by the at least one processor 51, and the instructions are executed by the at least one processor 51 to enable the at least one processor 51 to execute the picture advertisement recognition method.
Wherein the memory 52 and the processor 51 are coupled in a bus, which may comprise any number of interconnected buses and bridges, which couple one or more of the various circuits of the processor 51 and the memory 52 together. The bus may also connect various other circuits such as peripherals, voltage regulators, power management circuits, and the like, which are well known in the art, and therefore, will not be described any further herein. A bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or a plurality of elements, such as a plurality of receivers and transmitters, providing a means for communicating with various other apparatus over a transmission medium. The data processed by the processor 51 is transmitted over a wireless medium via an antenna, which further receives the data and transmits the data to the processor 51.
The processor 51 is responsible for managing the bus and general processing and may also provide various functions including timing, peripheral interfaces, voltage regulation, power management, and other control functions. And the memory 52 may be used to store data used by the processor 51 in performing operations.
The present invention further provides a computer-readable storage medium storing a computer program. The computer program realizes the above-described method embodiments when executed by a processor.
The scheme can obtain the word embedding characteristics in the picture advertisement, extract the picture embedding characteristics in the picture advertisement, construct a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics, identify whether the picture advertisement is the cheating picture advertisement according to the constructed two-classification model of the cheating picture advertisement, and improve the accuracy of identifying the cheating picture advertisement.
Furthermore, the above scheme can adopt a character recognition mode to obtain characters in the picture advertisement, and obtain the word embedding characteristics in the picture advertisement through a general semantic representation model mode, so that the advantage that the general semantic representation model can predict the forward and backward context of words, and the accuracy rate of the obtained word embedding characteristics can be improved.
Furthermore, the above scheme can extract the environmental scene information from the picture advertisement according to the preset environmental scene model, and extract the picture embedding feature in the picture advertisement by adopting the convolution neural network mode based on the environmental scene information, so that the advantage that the accuracy of the extracted picture embedding feature can be improved can be realized due to the fact that the environmental scene information in the picture advertisement and the information of the picture advertisement context can be reserved by the convolution neural network.
Furthermore, according to the scheme, the word embedding characteristics and the graph embedding characteristics can be subjected to differential operation and normalization operation, and the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation are used as a training input mode of the multi-mode deep semantic interaction neural network to construct the two classification models of the cheating picture advertisement.
Furthermore, the scheme can train and update the established binary model of the cheating picture advertisement through preset times of iteration, and the method has the advantage that the accuracy rate of identifying the cheating picture advertisement can be further improved.
In the several embodiments provided in the present invention, it should be understood that the disclosed system, apparatus and method may be implemented in other manners. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of a module or a unit is merely a logical division, and an actual implementation may have another division, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
Units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be substantially or partially implemented in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, a network device, or the like) or a processor (processor) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
The above description is only a part of the embodiments of the present invention, and not intended to limit the scope of the present invention, and all equivalent devices or equivalent processes performed by the present invention through the contents of the specification and the drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A picture advertisement identification method is characterized by comprising the following steps:
acquiring word embedding characteristics in the picture advertisement;
extracting graph embedding features in the picture advertisement;
constructing a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics;
and identifying whether the picture advertisement is a fraudulent picture advertisement or not according to the established binary classification model of the fraudulent picture advertisement.
2. The method for identifying picture advertisements as claimed in claim 1, wherein said obtaining word-embedded features in picture advertisements comprises:
the method comprises the steps of obtaining characters in the picture advertisement in a character recognition mode, and obtaining word embedding characteristics in the picture advertisement in a general semantic representation mode.
3. The method for identifying picture advertisements as claimed in claim 1, wherein the extracting of the picture embedding features in the picture advertisements comprises:
extracting environmental scene information from the picture advertisement according to a preset environmental scene model, and extracting the picture embedding characteristics in the picture advertisement by adopting a convolutional neural network mode based on the environmental scene information.
4. The picture advertisement recognition method of claim 1, wherein the building of the two-classification model of the fraudulent picture advertisement by means of a neural network manner of multi-modal deep semantic interaction based on the word embedding features and the picture embedding features comprises:
and carrying out differential operation and normalization operation on the word embedding characteristics and the graph embedding characteristics, and constructing a binary classification model of the cheating picture advertisement by taking the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation as a training input mode of a multi-mode deep semantic interaction neural network.
5. The method for identifying picture advertisements as claimed in claim 1, wherein after said identifying whether a picture advertisement is a fraudulent picture advertisement according to said constructed binary model of fraudulent picture advertisements, further comprising:
and training and updating the established binary model of the cheating picture advertisement through preset times of iteration.
6. An apparatus for recognizing a picture advertisement, comprising:
the device comprises an acquisition module, an extraction module, a construction module and an identification module;
the acquisition module is used for acquiring word embedding characteristics in the picture advertisement;
the extraction module is used for extracting the graph embedding characteristics in the picture advertisement;
the building module is used for building a two-classification model of the cheating picture advertisement by adopting a multi-mode deep semantic interaction neural network mode based on the word embedding characteristics and the picture embedding characteristics;
and the identification module is used for identifying whether the picture advertisement is a fraudulent picture advertisement according to the established binary classification model of the fraudulent picture advertisement.
7. The picture advertisement recognition device of claim 6, wherein the obtaining module is specifically configured to:
the method comprises the steps of obtaining characters in the picture advertisement in a character recognition mode, and obtaining word embedding characteristics in the picture advertisement in a general semantic representation mode.
8. The picture advertisement recognition device of claim 6, wherein the extraction module is specifically configured to:
extracting environmental scene information from the picture advertisement according to a preset environmental scene model, and extracting the picture embedding characteristics in the picture advertisement by adopting a convolutional neural network mode based on the environmental scene information.
9. The picture advertisement recognition device of claim 6, wherein the construction module is specifically configured to:
and carrying out differential operation and normalization operation on the word embedding characteristics and the graph embedding characteristics, and constructing a binary classification model of the cheating picture advertisement by taking the word embedding characteristics and the graph embedding characteristics subjected to the differential operation and the normalization operation as a training input mode of a multi-mode deep semantic interaction neural network.
10. The picture advertisement recognition apparatus according to claim 6, further comprising:
an update module;
and the updating module is used for training and updating the established binary model of the cheating picture advertisement through iteration of preset times.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010932388.4A CN112200598B (en) | 2020-09-08 | 2020-09-08 | Picture advertisement identification method and device and computer equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010932388.4A CN112200598B (en) | 2020-09-08 | 2020-09-08 | Picture advertisement identification method and device and computer equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112200598A true CN112200598A (en) | 2021-01-08 |
| CN112200598B CN112200598B (en) | 2022-02-15 |
Family
ID=74006483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010932388.4A Active CN112200598B (en) | 2020-09-08 | 2020-09-08 | Picture advertisement identification method and device and computer equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112200598B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112966774A (en) * | 2021-03-24 | 2021-06-15 | 黑龙江机智通智能科技有限公司 | Histopathology image classification method based on image Bert |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845352A (en) * | 2016-12-23 | 2017-06-13 | 北京旷视科技有限公司 | Pedestrian detection method and device |
| CN107563379A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | For the localization method to natural scene image Chinese version |
| CN107832765A (en) * | 2017-09-13 | 2018-03-23 | 百度在线网络技术（北京）有限公司 | Picture recognition to including word content and picture material |
| US20190354159A1 (en) * | 2017-10-29 | 2019-11-21 | Shanghai Cambricon Information Technology Co., Ltd | Convolutional operation device and method |
-
2020
- 2020-09-08 CN CN202010932388.4A patent/CN112200598B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106845352A (en) * | 2016-12-23 | 2017-06-13 | 北京旷视科技有限公司 | Pedestrian detection method and device |
| CN107563379A (en) * | 2017-09-02 | 2018-01-09 | 西安电子科技大学 | For the localization method to natural scene image Chinese version |
| CN107832765A (en) * | 2017-09-13 | 2018-03-23 | 百度在线网络技术（北京）有限公司 | Picture recognition to including word content and picture material |
| US20190354159A1 (en) * | 2017-10-29 | 2019-11-21 | Shanghai Cambricon Information Technology Co., Ltd | Convolutional operation device and method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112966774A (en) * | 2021-03-24 | 2021-06-15 | 黑龙江机智通智能科技有限公司 | Histopathology image classification method based on image Bert |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112200598B (en) | 2022-02-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111476284B (en) | Image recognition model training and image recognition method and device and electronic equipment | |
| CN110442841B (en) | Resume identification method and device, computer equipment and storage medium | |
| CN106649542B (en) | System and method for visual question answering | |
| CN111488826A (en) | Text recognition method and device, electronic equipment and storage medium | |
| CN102831200A (en) | Commodity propelling method and device based on image character recognition | |
| CN103646110A (en) | Natural person basic identity information matching method | |
| CN107908641A (en) | A kind of method and system for obtaining picture labeled data | |
| CN110110213B (en) | Method and device for mining user occupation, computer readable storage medium and terminal equipment | |
| CN102930048A (en) | Data abundance automatically found by semanteme and using reference and visual data | |
| CN112287069A (en) | Information retrieval method and device based on voice semantics and computer equipment | |
| CN113033438A (en) | Data feature learning method for modal imperfect alignment | |
| CN102741861B (en) | Cross the image identification system of complete dictionary based on cascade | |
| CN114972823A (en) | Data processing method, device, equipment and computer medium | |
| CN112200598B (en) | Picture advertisement identification method and device and computer equipment | |
| CN111709431B (en) | Instant translation method and device, computer equipment and storage medium | |
| WO2020175806A1 (en) | Device for recognizing characters and method for recognizing characters by means of same | |
| CN114780701A (en) | Automatic question-answer matching method, device, computer equipment and storage medium | |
| Islam et al. | BdSLW-11: Dataset of Bangladeshi sign language words for recognizing 11 daily useful BdSL words | |
| CN111126372B (en) | Logo region marking method and device in video and electronic equipment | |
| CN111125550A (en) | Interest point classification method, device, equipment and storage medium | |
| CN110852376B (en) | Method and system for identifying biological species | |
| KR20210010602A (en) | Apparatus and method for recognizing character | |
| CN111638792A (en) | AR effect presentation method and device, computer equipment and storage medium | |
| CN108536666A (en) | A kind of short text information extracting method and device | |
| CN111611812A (en) | Translating into braille |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20221130 Address after: Floor 14, Building 3, Chengying Center, No. 5 Yard, Laiguangying West Road, Chaoyang District, Beijing 100010 Patentee after: BEIJING NEXTDATA TIMES TECHNOLOGY Co.,Ltd. Patentee after: Shumei Tianxia (Beijing) Technology Co.,Ltd. Address before: 100012 14th floor, building 3, Wangjing Chengying center, Chaoyang District, Beijing Patentee before: BEIJING NEXTDATA TIMES TECHNOLOGY Co.,Ltd. |
|
| TR01 | Transfer of patent right |