CN116012924B - Face gallery construction method and device and computing equipment - Google Patents
Face gallery construction method and device and computing equipment Download PDFInfo
- Publication number
- CN116012924B CN116012924B CN202310095642.3A CN202310095642A CN116012924B CN 116012924 B CN116012924 B CN 116012924B CN 202310095642 A CN202310095642 A CN 202310095642A CN 116012924 B CN116012924 B CN 116012924B
- Authority
- CN
- China
- Prior art keywords
- picture
- face
- trusted
- people
- person
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The application discloses a face gallery construction method, a device and a computing device, wherein the method comprises the following steps: collecting at least one character picture matched with a character keyword; detecting whether at least one trusted people picture matched with the people keywords exists or not; if yes, screening at least one template picture from the at least one person picture according to the similarity of the faces of the at least one person picture and the at least one trusted person picture; if not, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; and constructing a face gallery according to at least one template picture and the character keywords. By means of the method, the face data are collected and marked, the whole process is completed manually only by participating in few confirmation works, human intervention is reduced, and the efficiency of building the face gallery is improved.
Description
Technical Field
The application relates to the technical field of computer vision, in particular to a face gallery construction method, a face gallery construction device and computing equipment.
Background
Face plays a vital role in the identification of the identity of a person in daily life, face detection and identification are two important tasks in the field of computer vision, and have been widely applied to a plurality of fields such as payment, security monitoring, video tracking and the like.
At present, the application of the face detection and recognition technology needs to rely on a large amount of face data with labels, the data acquisition work needs to consume a large amount of manpower, and the data acquisition method is used for acquiring, cleaning, labeling and the like, and is low in efficiency and high in cost.
Disclosure of Invention
In view of the foregoing, the present application is directed to a method, apparatus, and computing device for face gallery construction that overcomes or at least partially solves the foregoing problems.
According to one aspect of the application, a face gallery construction method is provided, and the method comprises the following steps:
collecting at least one character picture matched with a character keyword;
detecting whether at least one trusted people picture matched with the people keywords exists or not;
if yes, screening at least one template picture from the at least one person picture according to the similarity of the faces of the at least one person picture and the at least one trusted person picture;
If not, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture;
and constructing a face gallery according to at least one template picture and the character keywords.
Optionally, detecting whether there is at least one trusted people picture that matches the people keyword further includes:
and searching the matched trusted people pictures in the trusted people picture library by utilizing the people keywords to obtain trusted people picture searching results, and detecting whether the trusted people picture searching results are empty.
Optionally, the method further comprises:
searching in the information source according to the character keywords to obtain information search results;
extracting at least one trusted people picture corresponding to the people keyword from the information search result;
and storing at least one trusted character picture corresponding to the character keywords into a trusted character picture library.
Optionally, selecting at least one first candidate picture from the at least one person picture further comprises:
calculating cosine distances between face features of at least one person picture, clustering the at least one person picture according to the cosine distances, and screening at least one first candidate picture from the at least one person picture in the target class in the clustering result; wherein the number of people pictures in the target class is the largest.
Optionally, selecting at least one first candidate picture from the at least one person picture within the target class in the clustering result further comprises:
and filtering at least one person picture in the target class according to at least one filtering factor to obtain at least one first candidate picture.
Optionally, screening at least one template picture from the at least one persona picture according to the facial similarity between the at least one persona picture and the at least one trusted persona picture further includes:
determining at least one second candidate picture with the similarity of the face of the at least one trusted people picture meeting the preset condition in the at least one people picture;
and filtering the at least one second candidate picture according to the at least one filtering factor to obtain at least one template picture.
Optionally, extracting at least one trusted people picture corresponding to the people keyword from the information search result further includes:
extracting at least one candidate trusted people picture contained in the information search result, and filtering the at least one candidate trusted people picture according to at least one filtering factor; at least one trusted people picture is selected from the at least one candidate trusted people picture remaining after filtering.
Optionally, the at least one filter factor comprises: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio.
According to another aspect of the present application, there is provided a face gallery construction apparatus, including:
the acquisition module is suitable for acquiring at least one character picture matched with the character keywords;
the detection module is suitable for detecting whether at least one trusted character picture matched with the character keywords exists or not;
the screening module is suitable for screening at least one template picture from at least one character picture according to the similarity of the faces of the at least one character picture and the at least one trusted character picture if the at least one trusted character picture matched with the character keywords exists; or if at least one trusted person picture matched with the person keyword does not exist, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user can label at least one template picture in the at least one first candidate picture;
and the construction module is suitable for constructing a face drawing library according to at least one template picture and the character keywords.
Optionally, the detection module is further adapted to: and searching the matched trusted people pictures in the trusted people picture library by utilizing the people keywords to obtain trusted people picture searching results, and detecting whether the trusted people picture searching results are empty.
Optionally, the build module is further adapted to: searching in the information source according to the character keywords to obtain information search results; extracting at least one trusted people picture corresponding to the people keyword from the information search result; and storing at least one trusted character picture corresponding to the character keywords into a trusted character picture library.
Optionally, the screening module is further adapted to: calculating cosine distances between face features of at least one person picture, clustering the at least one person picture according to the cosine distances, and screening at least one first candidate picture from the at least one person picture in the target class in the clustering result; wherein the number of people pictures in the target class is the largest.
Optionally, the screening module is further adapted to: and filtering at least one person picture in the target class according to at least one filtering factor to obtain at least one first candidate picture.
Optionally, the screening module is further adapted to: determining at least one second candidate picture with the similarity of the face of the at least one trusted people picture meeting the preset condition in the at least one people picture;
And filtering the at least one second candidate picture according to the at least one filtering factor to obtain at least one template picture.
Optionally, the build module is further adapted to: extracting at least one candidate trusted people picture contained in the information search result, and filtering the at least one candidate trusted people picture according to at least one filtering factor; at least one trusted people picture is selected from the at least one candidate trusted people picture remaining after filtering.
Optionally, the at least one filter factor comprises: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio.
According to yet another aspect of the present application, there is provided a computing device comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the face gallery construction method.
According to still another aspect of the present application, there is provided a computer storage medium having at least one executable instruction stored therein, where the executable instruction causes a processor to perform operations corresponding to the face gallery construction method described above.
According to the face gallery construction method, the face gallery construction device and the computing equipment, at least one character picture matched with character keywords is acquired; detecting whether at least one trusted people picture matched with the people keywords exists or not; if yes, screening at least one template picture from the at least one person picture according to the similarity of the faces of the at least one person picture and the at least one trusted person picture; if not, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; and constructing a face gallery according to at least one template picture and the character keywords. By means of the method, the face data are collected and marked, the whole process is completed manually only by participating in few confirmation works, human intervention is reduced, and the efficiency of building the face gallery is improved.
The foregoing description is only an overview of the technical solutions of the present application, and may be implemented according to the content of the specification in order to make the technical means of the present application more clearly understood, and in order to make the above-mentioned and other objects, features and advantages of the present application more clearly understood, the following detailed description of the present application will be given.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the application. Also, like reference numerals are used to designate like parts throughout the figures. In the drawings:
fig. 1 shows a flowchart of a face gallery construction method provided in an embodiment of the present application;
fig. 2 shows a flowchart of a face gallery construction method according to another embodiment of the present application;
fig. 3 is a schematic structural diagram of a face gallery construction device according to another embodiment of the present application;
fig. 4 is a schematic structural diagram of a face gallery construction system according to another embodiment of the present disclosure;
fig. 5 shows a schematic structural diagram of a computing device according to an embodiment of the present application.
Detailed Description
Exemplary embodiments of the present application will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
Fig. 1 shows a flowchart of a face gallery construction method provided in an embodiment of the present application, where the method is applied to any device having computing capability. As shown in fig. 1, the method comprises the steps of:
step S110, at least one character picture matched with the character keywords is collected.
For example, a person keyword is used to search in a search engine to obtain at least one person picture matched with the person keyword, the person picture obtained in the step is a basic source for constructing a face gallery, and the searched person picture is a public picture resource for wide users.
Step S120, detecting whether there is at least one trusted people picture matching the people keyword.
The trusted people picture refers to a high-quality people picture with high matching degree with people, such as a news people picture extracted from a news report, a personal photo provided by a user, a people public picture in an information website and the like, and a corresponding relation between the trusted people picture and a people keyword can be stored in advance, so that the trusted people picture with the corresponding relation is queried according to the people keyword.
If at least one trusted people picture matched with the people keywords exists, executing step S130; if there is not at least one trusted people picture matching the people keyword, step S140 is performed.
Step S130, at least one template picture is selected from at least one person picture according to the similarity of the faces of the at least one person picture and at least one trusted person picture.
And if the trusted people pictures matched with the character keywords exist, screening template pictures from the individual people pictures through the trusted people pictures. Specifically, the face similarity between the person picture and the trusted person picture is calculated, and at least one template picture is selected from the person pictures according to the face similarity between each person picture and each trusted person picture, for example, the person picture with the face similarity exceeding the preset value is determined as the template picture.
Step S140, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user, so that the labeling user marks at least one template picture in the at least one first candidate picture.
If there is no trusted people picture matching the people keywords, the template picture cannot be screened from the various people pictures through the trusted people picture, and in order to ensure the high quality of the template picture, at least one first candidate picture is screened from the various people pictures and provided for the labeling user to label the at least one template picture from the labeling user.
And step S150, constructing a face gallery according to at least one template picture and the character keywords.
And aiming at all the template pictures, taking the character keywords as labeling information, and storing the labeling information and the template pictures in a face gallery in an associated manner.
And then, the face recognition can be realized through the face image library, wherein the character keywords corresponding to the face images in the face image library are the image labeling information, when the face images to be recognized are received, the face similarity between the face images to be recognized and the face images in the face image library is calculated, when the face similarity meets the preset condition, the corresponding face images are the face images matched with the face images to be recognized, and the character keywords corresponding to the face images are the character information, so that the face recognition is realized.
According to the face gallery construction method, at least one character picture matched with character keywords is collected; detecting whether at least one trusted people picture matched with the people keywords exists or not; if yes, screening at least one template picture from the at least one person picture according to the similarity of the faces of the at least one person picture and the at least one trusted person picture; if not, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; and constructing a face gallery according to at least one template picture and the character keywords. By the method, the collection and the labeling of the face data are realized, and only little confirmation work is needed to be participated in by the human in the whole process, so that the human intervention is reduced, and the efficiency of constructing the face gallery is improved.
Fig. 2 shows a flowchart of a face gallery construction method according to another embodiment of the present application, where the method is applied to any device having computing power. As shown in fig. 2, the method comprises the steps of:
step S210, at least one character picture matched with the character keywords is collected.
The person keywords can be person names, and then the person names are used for searching in public picture resources of the search engine, so that at least one person picture matched with the person keywords is obtained. Further, since the person may have a duplicate name, the person name and other feature keywords are used to form the person keywords for searching, so that the matching degree of the acquired person picture and the person is ensured. And in the subsequent process, screening and filtering each character picture obtained in the step to obtain the final character template picture in storage.
Step S220, searching the matched trusted people picture in the trusted people picture library by utilizing the people keywords to obtain a trusted people picture search result, and detecting whether the trusted people picture search result is empty.
The trusted people picture library stores the trusted people pictures of a plurality of people, and specifically stores the corresponding relation between the trusted people pictures and the people keywords, and then the trusted people pictures with the corresponding relation with the people keywords are searched in the trusted people picture library by using the people keywords, so that the trusted people pictures matched with the people keywords are obtained.
Specifically, the trusted people gallery is constructed by: searching in the information source according to the character keywords to obtain information search results; extracting at least one trusted people picture corresponding to the people keyword from the information search result; and storing at least one trusted character picture corresponding to the character keywords into a trusted character picture library.
Where the information source refers to a public information source available to the user, such as a search engine or a people data website, etc. The news report information of the person is obtained by searching in a search engine through the person keywords, the news report of the person is confirmed through text semantic analysis, the trusted person picture is extracted from the person pictures contained in the news report, and the trusted person picture is associated with the person keywords and is stored in a trusted person gallery.
In an alternative embodiment, extracting at least one trusted people picture corresponding to the people keyword from the information search result further includes: and extracting at least one candidate trusted people picture contained in the information search result, filtering the at least one candidate trusted people picture, and selecting at least one trusted people picture from at least one candidate trusted people picture remained after filtering. The at least one candidate trusted people picture contained in the information search result is all the people pictures contained in the information search result.
Specifically, the filter factors include: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio. The distance between the face and the center of the picture is used for evaluating whether the face in the picture is centered or not, specifically, the distance between the center position of the face (the center position of two eyes, the nose position and the like) and the center of the picture is smaller, the smaller the distance is, the closer the face area is to the center of the picture, the face is at the center of the picture when the distance between the face and the center of the picture is zero, and therefore, the smaller the distance between the face and the center of the picture is, the higher the quality of the picture of the person is considered; the face angle refers to the angle of face rotation when a photo is taken, and comprises three angles of a three-dimensional space, namely a left-right rotation angle, a head-up low head angle and a left-right deflection angle, wherein the smaller the face angle is, the higher the quality of a figure picture is considered, and the face angle can be obtained through a deep HeadPose algorithm; the face definition refers to the definition degree of a face in a picture, and the higher the face definition is, the higher the image quality of a person is considered, and the face definition can be obtained through a Bluedetection algorithm; the face area ratio refers to the ratio of the area of the face area to the whole picture area, and the larger the face area ratio is, the higher the quality of the figure picture is considered.
For example, a distance threshold is set, when the distance between the face and the center of the picture is smaller than the distance threshold, the face is judged to be centered, otherwise, the face is judged to be not centered, and candidate trusted people pictures with the non-centered faces are filtered; setting a face angle threshold, detecting the face angle of at least one candidate trusted people picture, and filtering out the candidate trusted people picture of which the face angle exceeds the face angle threshold; or setting a face definition threshold, detecting the face definition of at least one candidate trusted people picture, and filtering candidate trusted people pictures of which the face definition does not reach the face definition threshold; or setting a face area ratio threshold, detecting the face area ratio of at least one candidate trusted people picture, and filtering candidate trusted people pictures of which the face area ratio does not reach the face area ratio threshold; or, for each candidate trusted people picture, calculating the score according to at least two of the distance between the face and the picture center, the face angle, the face definition and the face area ratio, filtering the candidate trusted people picture with the score not meeting the preset condition, for example, respectively setting the weight value of each filtering factor, and calculating the weighted sum according to the weight value and the information value corresponding to the filtering factor to obtain the score. By the mode, the high-quality trusted people picture can be obtained.
In another alternative embodiment, selecting at least one trusted people picture from the at least one candidate trusted people picture remaining after filtering further comprises: and calculating cosine distance (namely face similarity) of the face features of at least one candidate trusted people picture remaining after filtering, carrying out clustering processing according to the face similarity, and determining at least one candidate trusted people picture contained in the class with the most data in the class as at least one trusted people picture. And clustering is carried out through the similarity of the faces, so that the matching degree of the characters and the trusted character pictures is ensured.
If the trusted people picture search result is not null, executing step S230; if the trusted people picture search result is empty, step S240 is performed.
Step S230, at least one second candidate picture with the face similarity of at least one trusted people picture contained in the trusted people picture search result meeting the preset condition is determined in at least one people picture; and filtering the at least one second candidate picture according to the at least one filtering factor to obtain at least one template picture.
If the trusted people picture search result contains at least one trusted people picture, the human face similarity between the at least one trusted people picture and each of the people pictures is calculated, and the corresponding people picture when the human face similarity meets the preset condition is determined to be the second candidate picture. The face similarity meeting the preset condition may be that the face similarity reaches a preset face similarity threshold, or that the face similarity is arranged in the front N (N is greater than 1) positions.
Specifically, a face detection algorithm is adopted to extract face features from the trusted people picture and the people picture, for example, retinaFace and AdaFace algorithms are adopted to perform face detection and extract face features. And solving cosine distances for the face features of the trusted character picture and the face features of the character picture to obtain the face similarity between the trusted character picture and the face features of the character picture.
And after determining at least one second candidate picture, filtering out part of pictures in the at least one second candidate picture according to at least one filtering factor to obtain at least one template picture. Wherein the at least one filter factor comprises: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio. For example, the distance threshold, the face angle threshold, the face definition threshold, and the face duty ratio threshold are set respectively for filtering, or the score is calculated according to at least two of the distance between the face and the center of the picture, the face angle, the face definition, and the face area duty ratio, and filtering is performed according to the score, and the detailed description of the filtering process according to the filtering factor is referred to above, and will not be repeated here. By the method, the picture with the face close-up and high quality can be screened out and used as the template picture.
In another alternative embodiment, after filtering at least one second candidate picture, the second candidate pictures remaining after filtering are ordered, and the second candidate picture arranged in the first M bits (M is greater than 1) is determined as the template picture.
Specifically, the ordering is based on face angle, face definition, and/or face area ratio. For example, respectively setting sorting weight values of three factors, calculating a weighted sum according to the sorting weight values and the information values to obtain sorting scores, and sorting all the second candidate pictures remained after filtering according to the sorting scores; for example, the priorities of the three factors are respectively set, and the sorting is performed according to the priorities and the information values, for example, the priority of the face definition is set to be higher than the priority of the face area ratio and the priority of the face area ratio is set to be higher than the priority of the face angle, the face definition of each second candidate picture is firstly compared, the higher the face definition is, the more front the sorting is, if the face definition of at least two second candidate pictures is consistent, the face area ratio is compared, the higher the face area ratio is, the more front the sorting is, and if the face area ratio of at least two second candidate pictures is consistent, the face angle is continuously compared, and the smaller the face angle is, the more front the sorting is.
Step S240, clustering is carried out on at least one person picture, and at least one first candidate picture is selected from at least one person picture in a target class in a clustering result; feeding back at least one first candidate picture to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; wherein the number of people pictures in the target class is the largest.
If the trusted people picture search result does not contain at least one trusted people picture, namely, the trusted people picture matched with the people keywords does not exist, the people face features of each person picture are extracted, cosine distances among the people face features of each person picture are calculated, each person picture is clustered according to the cosine distances, at least one first candidate picture is selected from the class with the largest number of contained person pictures, wherein the person pictures matched with the target person in the search result occupy a larger proportion based on the characteristics of a search engine, and the first candidate picture is selected from the class with the largest number of data in the class through clustering processing, so that the matching degree of the first candidate picture and the person can be ensured.
In an alternative embodiment, the selecting at least one first candidate picture from the at least one character picture within the target class in the clustering result further comprises: and filtering at least one person picture in the target class according to at least one filtering factor to obtain at least one first candidate picture. For example, the distance threshold, the face angle threshold, the face definition threshold, and the face duty ratio threshold are set respectively for filtering, or the score is calculated according to at least two of the distance between the face and the center of the picture, the face angle, the face definition, and the face area duty ratio, and filtering is performed according to the score, and the detailed description of the filtering process according to the filtering factor is referred to above, and will not be repeated here. By the method, the face close-up picture with clear front can be screened out.
Each first candidate picture is then provided to the labeling user, who labels at least one template picture from each first candidate picture.
Step S250, a face gallery is constructed according to at least one template picture and character keywords.
And storing each template picture and the corresponding character keywords thereof into a face drawing library. The character keywords are equivalent to the labeling information of the character pictures. Therefore, through the method of the embodiment of the application, the image data of the person are cleaned, screened and template-put in storage, so that the labeling process of the face picture is greatly simplified.
In an alternative embodiment, the picture data in the face gallery is automatically updated, for example, by monitoring a character list, and whenever a new character is added to the character list, a template picture of the character is obtained and added to the face gallery according to the method described above. By the method, the figure picture data can be automatically updated, so that the efficiency is improved, and the labor cost is reduced.
According to the face gallery construction method, trusted people pictures are extracted from trusted resources according to face keywords to construct a trusted people gallery, basic people pictures are collected according to the face keywords, if the trusted people pictures can be inquired according to the face keywords, the face similarity between the basic people pictures and the trusted people pictures is calculated, the basic people pictures with higher face similarity are clustered, and each basic picture contained in the largest class of pictures in the class is filtered according to the distance between the face and the picture center, the face angle, the face definition and/or the face duty ratio, so that a template picture for constructing the face gallery is finally obtained; if the trusted people pictures cannot be queried according to the face keywords, clustering the basic people pictures, screening out partial pictures, filtering according to the distance between the faces and the picture center, the face angle, the face definition and/or the face duty ratio to obtain candidate pictures to be confirmed manually, feeding the candidate pictures back to the labeling user, and confirming template pictures used for constructing a face gallery in the labeling user through the labeling user. By the method, aiming at a plurality of processes such as image acquisition, cleaning, screening and warehousing of the traditional face labeling work, the labeling process of the face images is greatly simplified by utilizing the computer image processing, face detection and recognition methods, the labor is only required to participate in few confirmation works in the whole process, the labor work and the cost are reduced, the construction efficiency of the face image library is improved, the automatic labeling, updating and maintenance of the face data are realized, meanwhile, the face images with clear front and high quality are screened out by various means to be used for constructing the face image library, and the quality of the images in the face image library can be ensured.
Fig. 3 shows a schematic structural diagram of a face gallery construction device according to another embodiment of the present application, where, as shown in fig. 3, the device includes:
the acquisition module 31 is adapted to acquire at least one character picture matched with the character keyword;
a detection module 32 adapted to detect whether there is at least one trusted people picture matching the people key;
a screening module 33, adapted to screen at least one template picture from at least one person picture according to the facial similarity between the at least one person picture and the at least one trusted person picture if there is at least one trusted person picture matching the person keyword; or if at least one trusted person picture matched with the person keyword does not exist, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user can label at least one template picture in the at least one first candidate picture;
the construction module 34 is adapted to construct a face gallery based on at least one template picture and the character keywords.
In an alternative embodiment, the detection module 32 is further adapted to:
And searching the matched trusted people pictures in the trusted people picture library by utilizing the people keywords to obtain trusted people picture searching results, and detecting whether the trusted people picture searching results are empty.
In an alternative embodiment, build module 34 is further adapted to: searching in the information source according to the character keywords to obtain information search results; extracting at least one trusted people picture corresponding to the people keyword from the information search result; and storing at least one trusted character picture corresponding to the character keywords into a trusted character picture library.
In an alternative embodiment, the screening module 33 is further adapted to: calculating cosine distances between face features of at least one person picture, clustering the at least one person picture according to the cosine distances, and screening at least one first candidate picture from the at least one person picture in the target class in the clustering result; wherein the number of people pictures in the target class is the largest.
In an alternative embodiment, the screening module 33 is further adapted to: and filtering at least one person picture in the target class according to at least one filtering factor to obtain at least one first candidate picture.
In an alternative embodiment, the screening module 33 is further adapted to: determining at least one second candidate picture with the similarity of the face of the at least one trusted people picture meeting the preset condition in the at least one people picture;
and filtering the at least one second candidate picture according to the at least one filtering factor to obtain at least one template picture.
In an alternative embodiment, build module 34 is further adapted to: extracting at least one candidate trusted people picture contained in the information search result, and filtering the at least one candidate trusted people picture according to at least one filtering factor; at least one trusted people picture is selected from the at least one candidate trusted people picture remaining after filtering.
In an alternative embodiment, the at least one filter factor comprises: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio.
According to the face gallery construction device, at least one character picture matched with a character keyword is collected; detecting whether at least one trusted people picture matched with the people keywords exists or not; if yes, screening at least one template picture from the at least one person picture according to the similarity of the faces of the at least one person picture and the at least one trusted person picture; if not, at least one first candidate picture is selected from the at least one person picture, and the at least one first candidate picture is fed back to the labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; and constructing a face gallery according to at least one template picture and the character keywords. By means of the method, the face data are collected and marked, the whole process is completed manually only by participating in few confirmation works, human intervention is reduced, and the efficiency of building the face gallery is improved.
Fig. 4 shows a schematic structural diagram of a face gallery construction system according to another embodiment of the present application, and as shown in fig. 4, the system includes a system back end 41 for automation processing and a system front end 42 for manual participation, where the system back end 41 includes: the functions of the collecting module 411, the detecting module 412, the screening module 413 and the constructing module 414 included in the back end of the system can be seen from the description in the foregoing embodiments, and will not be described herein again. The system front end 42 includes: the display module 421 and the receiving module 422.
The main functions of the system back-end 41 are as follows: and acquiring multi-source face data according to names and keywords of the characters, including news texts and pictures, acquiring a small amount of credible image data of the characters through text semantic screening, picture quality screening and face clustering screening, acquiring image data related to the specified characters in public resources by utilizing an image data acquisition algorithm at the rear end of the system, extracting face pictures in the image data by using a face detection algorithm, clustering and cleaning the face pictures, selecting proper face pictures as face picture templates through a quality evaluation algorithm, and adding the face pictures into a face picture library.
The system front end 42 is mainly used for confirming template pictures under the condition that trusted face pictures cannot be found, the system back end 41 provides candidate pictures and character keywords thereof to the system front end 42, and a display module 421 in the system front end 42 displays the candidate pictures and the character keywords thereof, and confirms whether the candidate pictures are template pictures or not and confirms labeling information by manpower; the receiving module 422 is configured to receive an image to be identified and provide the image to the system back end 41, and the system back end 41 determines a face picture matched with the image to be identified by comparing the face features in the image to be identified with the face features of the face images in the face gallery, where the character keywords of the face picture are character information corresponding to the image to be identified, so as to implement face identification.
The embodiment of the application provides a non-volatile computer storage medium, which stores at least one executable instruction, and the computer executable instruction can execute the face gallery construction method in any of the method embodiments.
FIG. 5 illustrates a schematic diagram of a computing device according to an embodiment of the present application, and the embodiments of the present application are not limited to a specific implementation of the computing device.
As shown in fig. 5, the computing device may include: a processor 502, a communication interface (Communications Interface) 504, a memory 506, and a communication bus 508.
Wherein: processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508. A communication interface 504 for communicating with network elements of other devices, such as clients or other servers. The processor 502 is configured to execute the program 510, and may specifically perform relevant steps in the foregoing embodiment of a face gallery construction method for a computing device.
In particular, program 510 may include program code including computer-operating instructions.
The processor 502 may be a central processing unit CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present application. The one or more processors included by the computing device may be the same type of processor, such as one or more CPUs; but may also be different types of processors such as one or more CPUs and one or more ASICs.
A memory 506 for storing a program 510. Memory 506 may comprise high-speed RAM memory or may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
The algorithms or displays presented herein are not inherently related to any particular computer, virtual system, or other apparatus. Various general-purpose systems may also be used with the teachings herein. The required structure for a construction of such a system is apparent from the description above. In addition, embodiments of the present application are not directed to any particular programming language. It will be appreciated that a variety of programming languages may be used to implement the teachings of the present application as described herein, and the above description of specific languages is provided for disclosure of preferred embodiments of the present application.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the present application may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the above description of exemplary embodiments of the application, various features of embodiments of the application are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the application and aiding in the understanding of one or more of the various application's aspects. However, the disclosed method should not be construed as reflecting the intention that: i.e., the claimed application requires more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of this application.
Those skilled in the art will appreciate that the modules in the apparatus of the embodiments may be adaptively changed and disposed in one or more apparatuses different from the embodiments. The modules or units or components of the embodiments may be combined into one module or unit or component and, furthermore, they may be divided into a plurality of sub-modules or sub-units or sub-components. Any combination of all features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or units of any method or apparatus so disclosed, may be used in combination, except insofar as at least some of such features and/or processes or units are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the present application and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.
Various component embodiments of the present application may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art will appreciate that some or all of the functionality of some or all of the components according to embodiments of the present application may be implemented in practice using a microprocessor or Digital Signal Processor (DSP). The present application may also be embodied as an apparatus or device program (e.g., computer program and computer program product) for performing a portion or all of the methods described herein. Such a program embodying the present application may be stored on a computer readable medium, or may have the form of one or more signals. Such signals may be downloaded from an internet website, provided on a carrier signal, or provided in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit the application, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The application may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The use of the words first, second, third, etc. do not denote any order. These words may be interpreted as names. The steps in the above embodiments should not be construed as limiting the order of execution unless specifically stated.
Claims (8)
1. The face gallery construction method is characterized by comprising the following steps:
collecting at least one character picture matched with a character keyword;
detecting whether at least one trusted people picture matched with the people keywords exists;
if yes, determining at least one second candidate picture with the face similarity meeting the preset condition in the at least one person picture, and filtering the at least one second candidate picture according to at least one filter factor to obtain at least one template picture;
if not, calculating the cosine distance between the face features of at least one person image, carrying out clustering processing on the at least one person image according to the cosine distance, screening at least one first candidate image from the at least one person image in the target class in the clustering result, and feeding the at least one first candidate image back to the labeling user so that the labeling user can label at least one template image in the at least one first candidate image; wherein the number of character pictures in the target class is the largest;
and constructing a face gallery according to the at least one template picture and the character keywords.
2. The method of claim 1, wherein the detecting whether there is at least one trusted people picture that matches the people keyword further comprises:
and searching the matched trusted people pictures in the trusted people gallery by utilizing the people keywords to obtain trusted people picture searching results, and detecting whether the trusted people picture searching results are empty.
3. The method according to claim 2, wherein the method further comprises:
searching in an information source according to the character keywords to obtain information search results;
extracting at least one trusted people picture corresponding to the people keyword from the information search result;
and storing at least one trusted character picture corresponding to the character keyword into a trusted character gallery.
4. The method of claim 3, wherein the extracting at least one trusted people picture corresponding to the people keyword from the information search results further comprises:
extracting at least one candidate trusted people picture contained in the information search result, and filtering the at least one candidate trusted people picture according to at least one filtering factor; at least one trusted people picture is selected from the at least one candidate trusted people picture remaining after filtering.
5. The method of claim 4, wherein the at least one filtering factor comprises: the distance between the face and the center of the picture, the face angle, the face definition and/or the face area ratio.
6. A face gallery construction apparatus, the apparatus comprising:
the acquisition module is suitable for acquiring at least one character picture matched with the character keywords;
the detection module is suitable for detecting whether at least one trusted character picture matched with the character keywords exists or not;
the screening module is suitable for determining at least one second candidate picture with the similarity of the face of the at least one trusted person picture meeting the preset condition in the at least one trusted person picture if at least one trusted person picture matched with the person keyword exists, and filtering the at least one second candidate picture according to at least one filtering factor to obtain at least one template picture; or if at least one trusted person picture matched with the person keyword does not exist, calculating cosine distance between face features of the at least one person picture, clustering the at least one person picture according to the cosine distance, screening at least one first candidate picture from the at least one person picture in a target class in a clustering result, and feeding the at least one first candidate picture back to a labeling user so that the labeling user marks at least one template picture in the at least one first candidate picture; wherein the number of character pictures in the target class is the largest;
And the construction module is suitable for constructing a face drawing library according to the at least one template picture and the character keywords.
7. A computing device, comprising: the device comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete communication with each other through the communication bus;
the memory is configured to store at least one executable instruction, where the executable instruction causes the processor to perform the operations corresponding to the face gallery construction method according to any one of claims 1 to 5.
8. A computer storage medium having stored therein at least one executable instruction for causing a processor to perform operations corresponding to the face gallery construction method of any one of claims 1-5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310095642.3A CN116012924B (en) | 2023-01-30 | 2023-01-30 | Face gallery construction method and device and computing equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310095642.3A CN116012924B (en) | 2023-01-30 | 2023-01-30 | Face gallery construction method and device and computing equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116012924A CN116012924A (en) | 2023-04-25 |
| CN116012924B true CN116012924B (en) | 2023-06-27 |
Family
ID=86030301
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310095642.3A Active CN116012924B (en) | 2023-01-30 | 2023-01-30 | Face gallery construction method and device and computing equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116012924B (en) |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102231149A (en) * | 2010-12-22 | 2011-11-02 | 辜进荣 | Method for searching visual information of mobile phone based on local characteristics |
| CN102622580A (en) * | 2012-02-20 | 2012-08-01 | 华焦宝 | Face detection and recognition method and system |
| CN102915372A (en) * | 2012-11-06 | 2013-02-06 | 成都理想境界科技有限公司 | Image retrieval method, device and system |
| CN102938065A (en) * | 2012-11-28 | 2013-02-20 | 北京旷视科技有限公司 | Facial feature extraction method and face recognition method based on large-scale image data |
| CN103218838A (en) * | 2013-05-11 | 2013-07-24 | 苏州华漫信息服务有限公司 | Automatic hair drawing method for human face cartoonlization |
| CN103514447A (en) * | 2013-10-18 | 2014-01-15 | 陕西高新实业有限公司 | Face recognition system based on smart phone |
| CN103903236A (en) * | 2014-03-10 | 2014-07-02 | 北京信息科技大学 | Method and device for reconstructing super-resolution facial image |
| CN104978550A (en) * | 2014-04-08 | 2015-10-14 | 上海骏聿数码科技有限公司 | Face recognition method and system based on large-scale face database |
| CN107977439A (en) * | 2017-12-07 | 2018-05-01 | 宁波亿拍客网络科技有限公司 | A kind of facial image base construction method |
| CN108874853A (en) * | 2018-04-02 | 2018-11-23 | 焦点科技股份有限公司 | A method of construction face picture library |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8995729B2 (en) * | 2011-08-30 | 2015-03-31 | The Mitre Corporation | Accelerated comparison using scores from coarse and fine matching processes |
| US20140105466A1 (en) * | 2012-10-16 | 2014-04-17 | Ocean Images UK Ltd. | Interactive photography system and method employing facial recognition |
-
2023
- 2023-01-30 CN CN202310095642.3A patent/CN116012924B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102231149A (en) * | 2010-12-22 | 2011-11-02 | 辜进荣 | Method for searching visual information of mobile phone based on local characteristics |
| CN102622580A (en) * | 2012-02-20 | 2012-08-01 | 华焦宝 | Face detection and recognition method and system |
| CN102915372A (en) * | 2012-11-06 | 2013-02-06 | 成都理想境界科技有限公司 | Image retrieval method, device and system |
| CN102938065A (en) * | 2012-11-28 | 2013-02-20 | 北京旷视科技有限公司 | Facial feature extraction method and face recognition method based on large-scale image data |
| CN103218838A (en) * | 2013-05-11 | 2013-07-24 | 苏州华漫信息服务有限公司 | Automatic hair drawing method for human face cartoonlization |
| CN103514447A (en) * | 2013-10-18 | 2014-01-15 | 陕西高新实业有限公司 | Face recognition system based on smart phone |
| CN103903236A (en) * | 2014-03-10 | 2014-07-02 | 北京信息科技大学 | Method and device for reconstructing super-resolution facial image |
| CN104978550A (en) * | 2014-04-08 | 2015-10-14 | 上海骏聿数码科技有限公司 | Face recognition method and system based on large-scale face database |
| CN107977439A (en) * | 2017-12-07 | 2018-05-01 | 宁波亿拍客网络科技有限公司 | A kind of facial image base construction method |
| CN108874853A (en) * | 2018-04-02 | 2018-11-23 | 焦点科技股份有限公司 | A method of construction face picture library |
Non-Patent Citations (1)
| Title |
|---|
| 基于差影法粗分割与多模板匹配的人脸检测;周彩霞 等;《计算机工程与设计》;第25卷(第10期);1648-1650 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116012924A (en) | 2023-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107169049B (en) | Application tag information generation method and device | |
| US9053357B2 (en) | Identifying images using face recognition | |
| Kumar et al. | The p-destre: A fully annotated dataset for pedestrian detection, tracking, and short/long-term re-identification from aerial devices | |
| US9218364B1 (en) | Monitoring an any-image labeling engine | |
| US9977951B2 (en) | Picture ranking method, and terminal | |
| RU2668717C1 (en) | Generation of marking of document images for training sample | |
| US9037600B1 (en) | Any-image labeling engine | |
| US20100211551A1 (en) | Method, system, and computer readable recording medium for filtering obscene contents | |
| EP3028184A1 (en) | Method and system for searching images | |
| CN102880729A (en) | Figure image retrieval method and device based on human face detection and recognition | |
| CN111368772B (en) | Identity recognition method, device, equipment and storage medium | |
| CN108734159B (en) | Method and system for detecting sensitive information in image | |
| CN111695453B (en) | Drawing recognition method and device and robot | |
| CN109241325B (en) | Large-scale face retrieval method and device based on depth features | |
| CN110263830B (en) | Image processing method, device and system and storage medium | |
| CN113963303A (en) | Image processing method, video recognition method, device, equipment and storage medium | |
| CN107203638B (en) | Monitoring video processing method, device and system | |
| CN112000834A (en) | Document processing method, device, system, electronic equipment and storage medium | |
| CN109783678B (en) | Image searching method and device | |
| CN111027622A (en) | Picture label generation method and device, computer equipment and storage medium | |
| CN116012924B (en) | Face gallery construction method and device and computing equipment | |
| Pandey et al. | A hierarchical clustering approach for image datasets | |
| Boteanu et al. | Hierarchical clustering pseudo-relevance feedback for social image search result diversification | |
| Bhatt et al. | Indian monuments classification using support vector machine | |
| CN111782945B (en) | Book searching method, computing device and storage 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |