CN115294504B - Marketing video auditing system based on AI - Google Patents
Marketing video auditing system based on AI Download PDFInfo
- Publication number
- CN115294504B CN115294504B CN202211186780.4A CN202211186780A CN115294504B CN 115294504 B CN115294504 B CN 115294504B CN 202211186780 A CN202211186780 A CN 202211186780A CN 115294504 B CN115294504 B CN 115294504B
- Authority
- CN
- China
- Prior art keywords
- module
- frame
- identification
- visual
- data
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a marketing video auditing system based on AI (artificial intelligence), which comprises a preprocessing module, an auditing model establishing module and a visual optimization module, wherein the preprocessing module is used for preprocessing a video to be audited, the auditing model establishing module is used for establishing a model based on video auditing, the visual optimization module is used for optimizing the audited video, the preprocessing module comprises a visual frame extracting module, a characteristic identification module, a single-frame data quantity identification module and a mathematical curve fitting module, the visual frame extracting module extracts video frame numbers which can be continuously identified by human eyes to serve as a unit of visual frame, the characteristic identification module is used for carrying out characteristic identification on the extracted visual frame, the single-frame data quantity identification module is used for carrying out data quantity identification on the visual frame number of each unit, and the mathematical curve fitting module is used for fitting the visual frame data quantity of each unit to a curve.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a marketing video auditing system based on AI.
Background
Along with the fire development of the short video industry, more and more people select to use the short video for marketing, but currently, the auditing mechanism of each platform is different, the situation that part of marketing words or patterns are deleted may occur, the marketing effect of the video is influenced, and even the video is off-shelf, and the manual auditing mode adopted at present is a conservative video auditing mode, but for different people, the auditing strength of the auditing mode can elastically change and is easy to miss, marketing companies need to audit the marketing video before uploading the video, so that the situation that the marketing video is audited by the platforms to cause reduction of weight and influence the marketing effect is prevented, and a patent with a publication number of CN107798304B discloses a method for fast auditing the video, wherein the shot boundary is secondarily detected by adopting HOG characteristic difference to eliminate wrong shot boundary benefits, and the shot boundary distance is less than 15 long shot boundary distance is removed by adopting the HOG difference to obtain M shot boundaries so as to fast audit the video, although the auditing speed can be improved to a certain extent, the generating process of the HOG characteristic noise is very sensitive, and the problem of extra noise is generated in the difference auditing process; therefore, it is necessary to design an AI-based marketing video auditing system with multi-directional filtering and feature extraction.
Disclosure of Invention
The invention aims to provide a marketing video auditing system based on AI, which aims to solve the problems in the background technology.
In order to solve the technical problems, the invention provides the following technical scheme: a marketing video auditing system based on AI comprises a preprocessing module, an auditing model establishing module and a visual optimization module, wherein the preprocessing module is used for preprocessing a video to be audited, the auditing model establishing module is used for establishing a model based on video auditing, the visual optimization module is used for further optimizing the audited video, the preprocessing module comprises a visual frame extracting module, a feature recognition module, a single-frame data quantity recognition module and a mathematical curve fitting module, the visual frame extracting module extracts the number of video frames which can be continuously recognized by human eyes to serve as a unit of visual frame, the feature recognition module is used for performing feature recognition on the extracted visual frame, the single-frame data quantity recognition module is used for performing data quantity recognition on the visual frame of each unit, and the mathematical curve fitting module is used for fitting the data quantity of the visual frame of each unit to be a curve.
According to the technical scheme, the auditing model establishing module comprises a feature migration module, a convolution layering module, an original domain accumulation module and an identification domain selection module, wherein the feature migration module is used for model migration of features identified in the preprocessing process, the convolution layering module is used for convolution layering according to feature types, the original domain accumulation module is used for providing multiple features of an original data model, and the identification domain selection module is used for selecting an identification domain matched with the original data model according to a preprocessing result.
According to the technical scheme, the visual optimization module comprises a multidirectional filtering module, a single-frame confidence coefficient calculating module and a feedback module, the multidirectional filtering module is used for carrying out multidirectional filtering on the video to be examined, the single-frame confidence coefficient calculating module is used for carrying out confidence coefficient calculation on the split single-frame image, and the feedback module is used for carrying out feedback based on a time point on the video to be examined according to the confidence coefficient.
According to the technical scheme, in the system, the operation method of the preprocessing module comprises the following steps:
determining a visual frame, and extracting a multi-unit visual frame from the video;
performing feature recognition on the extracted multi-unit visual frame, wherein the feature recognition content comprises direct features, semantic features and repeated features;
carrying out data quantity identification statistics on the visual frame of each unit, and fitting a data quantity change curve by taking time as a reference;
and screening out the visual frames needing further examination by combining the identification characteristics and the data volume change curve.
According to the technical scheme, the screening method of the visual frames needing further auditing comprises the following steps:
after the feature recognition is carried out on the visual frame of a unit, a feature matrix of the visual frame of the unit is generated
；
Performing weighted average on elements in the feature matrix to generate a weight value of the unit visual frame
；
Setting an audit weight threshold
When it comes to
Time marks the time point of the unit visual frame and takes the unit visual frame as a candidate auditing visual frame;
amount of data contained in each unit visual frame
Identifying and outputting, and fitting time points
Graph with data amount, wherein
Is a set of data arrays of data amount in bytes, and the change of data amount contained in unit visual frame is obeyed by time point
The curve change of (c);
calculating the change trend of the visual frame data of each adjacent time point, and calculating the data change degree
；
And judging whether the data change is abnormal or not according to the AI, and marking and uploading abnormal points.
According to the technical scheme, the data change degree
The calculation method comprises the following steps:
in the formula, a whole video channel is represented
After the extraction of the identification frames of the units, the average data amount of each unit identification frame,
representing the variance value of all unit identification frames, directly reflecting each segmentA degree of data change for the frame is identified, wherein,
is as follows
The unit identifies the amount of data of the frame, in bytes,
the number of frames is identified for the total decimation,
the larger the data size is, the more obvious the change of the identification frame data of the section is, and the larger the fluctuation is;
and after the identification is finished, judging whether the data change is normal according to the AI, and carrying out time point marking on the abnormal identification frame and uploading.
According to the technical scheme, after the abnormal identification frame is uploaded, the verification model is used for further verification, and the establishment method of the verification model comprises the following steps:
step S1: establishing a characteristic original domain, specifically a characteristic original domain containing all characteristic ranges of the identification frame;
step S2: feature matrix
Carrying out convolution layering with the characteristic original domain, and establishing an auditing model with characteristic layering;
and step S3: after each audit, performing original domain accumulation on the identification of the features in the audit content, and storing the identification in a feature matrix form;
and step S4: after the verification is finished, selecting the identification domains with the corresponding layers for identification operation, and after the identification is passed, canceling the mark of the identification frame content, otherwise, continuously outputting.
According to the above technical solution, the operation method of the vision optimization module further comprises the steps of:
performing multi-directional filtering based on feature layering, wherein the filtering degree is related to the feature layer number of the visual frame;
calculating single frame confidence of unit visual frame according to characteristic parameters
Wherein the characteristic parameters comprise the total characteristic convolution layer number
Number of characteristic layers for filtration
Conversion coefficient of
。
According to the technical scheme, the single frame confidence degree
The calculation method comprises the following steps:
in the formula (I), the compound is shown in the specification,
the feature recognition depth after filtering is represented,
the proportion of the identification depth to the total characteristic convolution layer number is represented, the larger the proportion of the identification depth to the total characteristic convolution layer number is, the less the characteristic redundancy is represented, the higher the safety factor of a single identification frame is, and specifically, the converted coefficient is used
A converted single frame confidence representation;
after single-frame confidence coefficient calculation, marking the visual frames in the auditing model, performing danger feedback on the visual frames with the confidence coefficient lower than the average confidence coefficient, sending the visual frames to a manual auditing end for further auditing, and performing repeated model auditing on the visual frames with the confidence coefficient higher than the average confidence coefficient to realize multiple screening.
Compared with the prior art, the invention has the following beneficial effects: according to the invention, the visual frame extraction module is arranged, so that the video frame number which can be smoothly identified by human eyes is extracted as the visual frame, the AI operation complexity is reduced while the manual review is simulated, and the subsequent curve fitting is facilitated; the data volume in the identification frame of each unit is identified by arranging a single-frame data volume identification module, and a curve change image is fitted by using a time point for auxiliary identification; the feature matrix and the original feature domain are subjected to convolution layering through the convolution layering module, the auditing precision is improved, the original domain is updated, the confidence coefficient of a single identification frame is calculated through the single-frame confidence coefficient calculating module, the content of the identification frame is further screened, the part with abnormal numerical values is submitted to a manual place for further auditing, and the auditing efficiency is improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a schematic block diagram of the system of the present invention;
FIG. 2 is a diagram illustrating a single frame data variation curve according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-2, the present invention provides the following technical solutions: a marketing video auditing system based on AI comprises a preprocessing module, an auditing model establishing module and a visual optimization module, wherein the preprocessing module is used for preprocessing a video to be audited, the auditing model establishing module is used for establishing a model based on video auditing, the visual optimization module is used for further optimizing the audited video, the preprocessing module comprises a visual frame extracting module, a feature recognition module, a single-frame data quantity recognition module and a mathematical curve fitting module, the visual frame extracting module extracts the number of video frames which can be continuously recognized by human eyes to serve as a unit of visual frame, the feature recognition module is used for carrying out feature recognition on the extracted visual frame, the single-frame data quantity recognition module is used for carrying out data quantity recognition on the visual frame of each unit, and the mathematical curve fitting module is used for fitting the data quantity of the visual frame of each unit to be a curve.
The auditing model establishing module comprises a feature migration module, a convolution layering module, an original domain accumulation module and an identification domain selection module, wherein the feature migration module is used for performing model migration on the identified features in the preprocessing process, the convolution layering module is used for performing convolution layering according to feature types, the original domain accumulation module is used for providing multiple features of an original data model, and the identification domain selection module is used for selecting an identification domain matched with the original data model according to the preprocessing result.
The visual optimization module comprises a multidirectional filtering module, a single-frame confidence coefficient calculating module and a feedback module, wherein the multidirectional filtering module is used for carrying out multidirectional filtering on the video to be examined, the single-frame confidence coefficient calculating module is used for carrying out confidence coefficient calculation on the split single-frame image, and the feedback module is used for carrying out feedback based on time points on the video to be examined according to the confidence coefficient.
In the system, the operation method of the preprocessing module comprises the following steps:
determining a visual frame, and extracting multi-unit visual frames from the video; the visual frames are extracted by taking the number of video frames which can be smoothly identified by human eyes as a unit, so that the consistency of video audit is ensured;
performing feature recognition on the extracted multi-unit visual frame, wherein the feature recognition content comprises direct features, semantic features and repeated features;
carrying out data quantity identification statistics on the visual frame of each unit, and fitting a data quantity change curve by taking time as a reference;
and screening out the visual frames needing further examination by combining the identification characteristics and the data volume change curve.
The screening method of the visual frames needing further auditing comprises the following steps:
after the feature recognition is carried out on the visual frame of a unit, a feature matrix of the visual frame of the unit is generated
(ii) a After the visual frame is subjected to feature recognition, generating
Line of
A feature matrix of columns, the matrix containing all features of a single visual frame;
performing weighted average on elements in the feature matrix to generate a weight value of the unit visual frame
(ii) a After elements in the feature matrix are weighted and averaged, screening out elements which are least matched and excessively matched with the identification content and the features, generating a weight value of the visual frame, and feeding back an audit weight of the visual frame;
setting an audit weight threshold
When is coming into contact with
Time marking the time point of the unit visual frame, and using the time point as a candidate audit visual frame; auditing weight threshold
The big data is obtained by filtering according to the weight value of the identification content in the historical audit record, and is a dynamic value, the audit content is different every time, and the weight threshold is differentThe value also changes;
amount of data contained in each unit visual frame
Recognizing and outputting, and fitting out time points
Graph with data amount, wherein
Is a set of data arrays of data amount in bytes, and the change of data amount contained in unit visual frame is obeyed by time point
The curve change of (c); counting the data quantity contained in each identification frame, fitting a data change curve with the time point, recording the data flow in a set form, and judging the time point with abnormal data quantity by combining curve data;
calculating the change trend of the visual frame data of each adjacent time point, and calculating the data change degree
；
And judging whether the data change is abnormal or not according to the AI, and marking and uploading abnormal points.
Degree of data change
The calculating method comprises the following steps:
in the formula (I), the compound is shown in the specification,
representing a whole video channel
After the extraction of the identification frames of the units, the average data amount of each unit identification frame,
the variance value of all unit identification frames is expressed, and the data change degree of each identification frame is directly reflected, wherein,
is as follows
The unit identifies the amount of data of the frame, in bytes,
the number of frames is identified for the total decimation,
the larger the data size is, the more obvious the change of the identification frame data of the section is, and the larger the fluctuation is;
and after the identification is finished, judging whether the data change is normal according to the AI, and carrying out time point marking on the abnormal identification frame and uploading.
After the abnormal recognition frame is uploaded, the auditing model is used for further auditing, and the establishing method of the auditing model comprises the following steps:
step S1: establishing a characteristic original domain, specifically a characteristic original domain containing all characteristic ranges of the identification frame; the original features contain larger redundancy, an original feature domain is established after classification is carried out through a classifier, effective features are screened out according to AI and put into the domain, and the complexity of the feature original domain is reduced;
step S2: feature matrix
Carrying out convolution layering on the feature original domain, and establishing an auditing model with feature layering; after convolution of the feature matrix and the elements in the feature original domain, the feature matrix containing a plurality of directions is generatedLayering vectors, accessing the vectors into an audit model, matching most features in a visual frame, uploading the unmatched parts and updating an original domain to complete accumulation of the original domain, wherein the more the original domain is accumulated, the more accurate the feature identification is;
and step S3: after each audit, performing original domain accumulation on the identification of the features in the audit content, and storing the identification in a feature matrix form;
and step S4: and after the auditing is finished, selecting the identification domain with the corresponding layer number for identification operation, and after the identification is passed, identifying the frame content canceling mark, otherwise, continuously outputting.
The method for operating the vision optimization module further comprises the following steps:
performing multi-directional filtering based on feature layering, wherein the filtering degree is related to the feature layer number of the visual frame; the more the number of the characteristic layers is, the richer the video content is represented, the more the hidden characteristics are buried, and the more the common characteristics needing to be filtered are;
calculating single frame confidence of unit visual frame according to characteristic parameters
Wherein the characteristic parameter comprises the total characteristic convolution layer number
Number of characteristic layers for filtration
Conversion coefficient of
。
Single frame confidence
The calculation method comprises the following steps:
in the formula (I), the compound is shown in the specification,
the feature recognition depth after filtering is represented,
the proportion of the identification depth to the total characteristic convolution layer number is represented, the larger the proportion of the identification depth to the total characteristic convolution layer number is, the less the characteristic redundancy is represented, the higher the safety factor of a single identification frame is, and specifically, the converted coefficient is used
A converted single frame confidence representation;
after single-frame confidence coefficient calculation, marking the visual frames in the auditing model, performing danger feedback on the visual frames with the confidence coefficient lower than the average confidence coefficient, sending the visual frames to a manual auditing end for further auditing, and performing repeated model auditing on the visual frames with the confidence coefficient higher than the average confidence coefficient to realize multiple screening.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus.
Finally, it should be noted that: although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that changes may be made in the embodiments and/or equivalents thereof without departing from the spirit and scope of the invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (7)
1. A marketing video auditing system based on AI comprises a preprocessing module, an auditing model establishing module and a visual optimization module, and is characterized in that: the system comprises a preprocessing module, an audit model establishing module, a vision optimizing module, a characteristic identification module, a single-frame data quantity identification module and a mathematical curve fitting module, wherein the preprocessing module is used for preprocessing a video to be audited, the audit model establishing module is used for establishing a model based on video audit, the vision optimizing module is used for further optimizing the audited video, the preprocessing module comprises a vision frame extracting module, a characteristic identification module, a single-frame data quantity identification module and a mathematical curve fitting module, the vision frame extracting module extracts video frame numbers which can be continuously identified by human eyes to serve as a unit of vision frame, the characteristic identification module is used for carrying out characteristic identification on the extracted vision frame, the single-frame data quantity identification module is used for carrying out data quantity identification on the vision frame of each unit, and the mathematical curve fitting module is used for fitting the data quantity of the vision frame of each unit to a curve;
the auditing model establishing module comprises a feature migration module, a convolution layering module, an original domain accumulation module and an identification domain selection module, wherein the feature migration module is used for carrying out model migration on features identified in the preprocessing process, the convolution layering module is used for carrying out convolution layering according to feature types identified in the preprocessing process, the original domain accumulation module is used for providing multiple features of an original data model, and the identification domain selection module is used for selecting an identification domain matched with the original data model according to a preprocessing result; the visual optimization module comprises a multidirectional filtering module, a single-frame confidence coefficient calculation module and a feedback module, wherein the multidirectional filtering module is used for carrying out multidirectional filtering on the video to be reviewed, the single-frame confidence coefficient calculation module is used for carrying out confidence coefficient calculation on the split single-frame image, and the feedback module is used for carrying out feedback based on a time point on the video to be reviewed according to the confidence coefficient.
2. An AI-based marketing video audit system according to claim 1 wherein: the operation method of the pretreatment module comprises the following steps:
determining a visual frame, and extracting a multi-unit visual frame from the video;
performing feature recognition on the extracted multi-unit visual frame, wherein the feature recognition content comprises direct features, semantic features and repeated features;
carrying out data quantity identification statistics on the visual frame of each unit, and fitting a data quantity change curve by taking time as a reference;
and screening out the visual frames needing further examination by combining the identification characteristics and the data volume change curve.
3. The AI-based marketing video audit system according to claim 2 wherein: the screening method of the visual frames needing further auditing comprises the following steps:
after the feature recognition is carried out on the visual frame of one unit, a feature matrix of the visual frame of the unit is generated
；
Performing weighted average on elements in the feature matrix to generate a weight value of the unit visual frame
；
Setting an audit weight threshold
When is coming into contact with
Time marks the time point of the unit visual frame and takes the unit visual frame as a candidate auditing visual frame;
amount of data contained in each unit visual frame
Identifying and outputting, and fitting time points
Graph with data amount, wherein
Is a set of arrays of data volumes,
is a first
The unit identifies the amount of data of the frame, in bytes,
identifying the number of frames in bytes for the total extraction, the change of the data amount contained in the unit visual frame being obeyed by the time point
The curve change of (c);
calculating the change trend of the visual frame data of each adjacent time point, and calculating the data change degree
；
And judging whether the data change is abnormal or not according to the AI, and marking and uploading abnormal points.
4. An AI-based marketing video audit system according to claim 3 wherein: degree of change of the data
The calculating method comprises the following steps:
in the formula (I), the compound is shown in the specification,
representing a whole video channel
After the extraction of the identification frames of each unit, the average data amount of each unit identification frame,
the variance value of all unit identification frames is expressed, and the data change degree of each identification frame is directly reflected, wherein,
is as follows
The unit identifies the amount of data of the frame, in bytes,
for the total number of extracted identification frames,
the larger the data size is, the more obvious the change of the identification frame data of the section is, and the larger the fluctuation is;
and after the identification is finished, judging whether the data change is normal according to the AI, and carrying out time point marking on the abnormal identification frame and uploading.
5. The AI-based marketing video audit system according to claim 4 wherein: after the abnormal identification frame is uploaded, further auditing is carried out by an auditing model, and the establishing method of the auditing model comprises the following steps:
step S1: establishing a characteristic original domain, specifically a characteristic original domain containing all characteristic ranges of the identification frame;
step S2: feature matrix
Carrying out convolution layering on the feature original domain, and establishing an auditing model with feature layering;
and step S3: after each audit, performing original domain accumulation on the identification of the features in the audit content, and storing the identification in a feature matrix form;
and step S4: and after the auditing is finished, selecting the identification domain with the corresponding layer number for identification operation, and after the identification is passed, identifying the frame content canceling mark, otherwise, continuously outputting.
6. The AI-based marketing video audit system according to claim 5 wherein: the method for operating the vision optimization module further comprises the following steps:
performing multi-directional filtering based on feature layering;
calculating single-frame confidence coefficient of unit visual frame according to characteristic parameters
Wherein the characteristic parameter comprises the total characteristic convolution layer number
Number of characteristic layers for filtration
Conversion coefficient of
。
7. The AI-based marketing video audit system according to claim 6 wherein: the single frame confidence
The calculating method comprises the following steps:
in the formula (I), the compound is shown in the specification,
the feature recognition depth after filtering is represented,
the proportion of the identification depth to the total characteristic convolution layer number is represented, the larger the proportion of the identification depth to the total characteristic convolution layer number is, the less the characteristic redundancy is represented, the higher the safety factor of a single identification frame is, and specifically, the converted coefficient is used
A converted single frame confidence representation; after single-frame confidence calculation, marking the visual frames in the auditing model, performing danger feedback on the visual frames with the confidence lower than the average confidence, sending the visual frames to a manual auditing end for further auditing, and performing repeated model auditing on the visual frames with the confidence higher than the average confidence to realize multiple screening.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211186780.4A CN115294504B (en) | 2022-09-28 | 2022-09-28 | Marketing video auditing system based on AI |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211186780.4A CN115294504B (en) | 2022-09-28 | 2022-09-28 | Marketing video auditing system based on AI |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115294504A CN115294504A (en) | 2022-11-04 |
| CN115294504B true CN115294504B (en) | 2023-01-03 |
Family
ID=83834631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211186780.4A Active CN115294504B (en) | 2022-09-28 | 2022-09-28 | Marketing video auditing system based on AI |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115294504B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108124191A (en) * | 2017-12-22 | 2018-06-05 | 北京百度网讯科技有限公司 | A kind of video reviewing method, device and server |
| CN113850162A (en) * | 2021-09-10 | 2021-12-28 | 北京百度网讯科技有限公司 | Video auditing method and device and electronic equipment |
| CN113887432A (en) * | 2021-09-30 | 2022-01-04 | 瑞森网安(福建)信息科技有限公司 | Video auditing method and system |
| CN114385837A (en) * | 2021-12-20 | 2022-04-22 | 山东智驱力人工智能科技有限公司 | Automatic media content detection and verification method and system |
| CN114998782A (en) * | 2022-05-18 | 2022-09-02 | 平安科技（深圳）有限公司 | Scene classification method and device of face-check video, electronic equipment and storage medium |
| CN115049953A (en) * | 2022-05-09 | 2022-09-13 | 中移（杭州）信息技术有限公司 | Video processing method, device, equipment and computer readable storage medium |
| CN115100561A (en) * | 2022-06-08 | 2022-09-23 | 新疆大学 | Intelligent auditing method and system for bad content of network short video |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA3216076A1 (en) * | 2015-07-16 | 2017-01-19 | Inscape Data, Inc. | Detection of common media segments |
-
2022
- 2022-09-28 CN CN202211186780.4A patent/CN115294504B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108124191A (en) * | 2017-12-22 | 2018-06-05 | 北京百度网讯科技有限公司 | A kind of video reviewing method, device and server |
| CN113850162A (en) * | 2021-09-10 | 2021-12-28 | 北京百度网讯科技有限公司 | Video auditing method and device and electronic equipment |
| CN113887432A (en) * | 2021-09-30 | 2022-01-04 | 瑞森网安(福建)信息科技有限公司 | Video auditing method and system |
| CN114385837A (en) * | 2021-12-20 | 2022-04-22 | 山东智驱力人工智能科技有限公司 | Automatic media content detection and verification method and system |
| CN115049953A (en) * | 2022-05-09 | 2022-09-13 | 中移（杭州）信息技术有限公司 | Video processing method, device, equipment and computer readable storage medium |
| CN114998782A (en) * | 2022-05-18 | 2022-09-02 | 平安科技（深圳）有限公司 | Scene classification method and device of face-check video, electronic equipment and storage medium |
| CN115100561A (en) * | 2022-06-08 | 2022-09-23 | 新疆大学 | Intelligent auditing method and system for bad content of network short video |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115294504A (en) | 2022-11-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110490239B (en) | Training method, quality classification method, device and equipment of image quality control network | |
| CN112819604A (en) | Personal credit evaluation method and system based on fusion neural network feature mining | |
| CN107291822A (en) | The problem of based on deep learning disaggregated model training method, sorting technique and device | |
| CN107909299A (en) | People hinders Claims Resolution data risk checking method and system | |
| CN108257122A (en) | Paper sheet defect detection method, device and server based on machine vision | |
| CN111079941B (en) | Credit information processing method, credit information processing system, terminal and storage medium | |
| CN112906500B (en) | Facial expression recognition method and system based on deep privilege network | |
| CN112819093A (en) | Man-machine asynchronous recognition method based on small data set and convolutional neural network | |
| CN117455417B (en) | Automatic iterative optimization method and system for intelligent wind control approval strategy | |
| CN112164033A (en) | Abnormal feature editing-based method for detecting surface defects of counternetwork texture | |
| CN116524607A (en) | Human face fake clue detection method based on federal residual error | |
| CN115294504B (en) | Marketing video auditing system based on AI | |
| CN110059126B (en) | LKJ abnormal value data-based complex correlation network analysis method and system | |
| CN112889075B (en) | Improved predictive performance using asymmetric hyperbolic tangent activation function | |
| KR102525491B1 (en) | Method of providing structure damage detection report | |
| CN111666872A (en) | Efficient behavior identification method under data imbalance | |
| CN115187266B (en) | Credit card fraud detection method and system based on memory variation self-coding model | |
| CN116467930A (en) | Transformer-based structured data general modeling method | |
| CN116363344A (en) | Wheat spike counting method and system based on improved YOLOv5s | |
| CN114299328A (en) | Environment self-adaptive sensing small sample endangered animal detection method and system | |
| CN118296389B (en) | Construction and evaluation method of data index model | |
| CN114240929A (en) | Color difference detection method and device | |
| KR20220074327A (en) | Loan regular auditing system using artificia intellicence | |
| CN116259091B (en) | Method and device for detecting silent living body | |
| CN113705729B (en) | Garbage classification model modeling method, garbage classification device and medium |
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 | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: Room 11, room 9, floor 2, building 18, office R & D building, Huagong science and Technology Park, 33 Tangxun Hubei Road, Donghu New Technology Development Zone, Wuhan City, Hubei Province, 430000 Applicant after: Wuhan Dangxia time culture creative Co.,Ltd. Address before: 430223 Floor 2, R&D Building B, No. 1 Modern Service Industry Base, Huazhong University of Science and Technology Science Park, Donghu Hi tech Zone, Wuhan, Hubei Province Applicant before: Wuhan Dangxia time culture creative Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP01 | Change in the name or title of a patent holder | ||
| CP01 | Change in the name or title of a patent holder |
Address after: Room 11, room 9, floor 2, building 18, office R & D building, Huagong science and Technology Park, 33 Tangxun Hubei Road, Donghu New Technology Development Zone, Wuhan City, Hubei Province, 430000 Patentee after: Wuhan Dangxia Technology Co.,Ltd. Address before: Room 11, room 9, floor 2, building 18, office R & D building, Huagong science and Technology Park, 33 Tangxun Hubei Road, Donghu New Technology Development Zone, Wuhan City, Hubei Province, 430000 Patentee before: Wuhan Dangxia time culture creative Co.,Ltd. |