CN108052555A - A kind of photo classification method and system - Google Patents
A kind of photo classification method and system Download PDFInfo
- Publication number
- CN108052555A CN108052555A CN201711231374.4A CN201711231374A CN108052555A CN 108052555 A CN108052555 A CN 108052555A CN 201711231374 A CN201711231374 A CN 201711231374A CN 108052555 A CN108052555 A CN 108052555A
- Authority
- CN
- China
- Prior art keywords
- photo
- user
- primary election
- base
- group
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/5866—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using information manually generated, e.g. tags, keywords, comments, manually generated location and time information
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Library & Information Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses a kind of photo classification method and systems, belong to image identification technical field.Method includes:Multiple base photos are carried out by similarity calculation with every reference picture based on default similarity algorithm respectively, and an at least base photo of the similarity value higher than default similarity threshold is stored as primary election photo group to database;Based on user to the annotation results of every base photo in the primary election photo group, judge every base photo whether with the reference picture for same person;If so, the base photo is copied to the catalogue where the reference picture.The present invention can meet the requirement to the accuracy rate of photo classification, and having to meet reduces cost of labor demand, can realize that low cost, the efficient photo to magnanimity are classified.
Description
Technical field
The present invention relates to image identification technical field, more particularly to a kind of photo classification method and system.
Background technology
Recognition of face (Facial Recognition) is exactly the face-image that user is obtained by video capture device,
Face recognition algorithms is recycled to carry out calculating analysis, and then and own database to the face position, shape of face and angle of its face
In existing model be compared, after judge the true identity of user.At present, face recognition technology deploy to ensure effective monitoring and control of illegal activities in monitoring,
The fields such as public security photo search system, gate inhibition's discrepancy, identification, which realize, to be widely applied.
The core technology of recognition of face is face recognition algorithms.The recognition success rate for influencing face recognition algorithms is subject to very much
The influence of factor, such as:Photo looks after variation, the change of age of personage, environmental change, the variation of personage's bodily form, personage's posture
Variation, picture quality etc. can all influence the accuracy of face recognition algorithms.
In the implementation of the present invention, inventor has found, based on same person in different scenes (different illumination, ring
Border, age, expression and attachment (cap, glasses)) under photo carry out face recognition algorithms training, recognition of face can be improved
Recognition success rate of the algorithm under complex environment.And this mode needs first to complete the classification to photo, i.e., in units of individual
Classify to photo.Since photo is magnanimity, it is completely dependent on manually carrying out classifying needing to expend substantial amounts of manpower, financial resources;
It due to the limited accuracy of algorithm, is completely dependent on algorithm and classifies, it is impossible to realize the precision of classificating requirement.
Therefore, there is an urgent need for a kind of high efficiency, low cost, simple, easy and photo classification methods with high accuracy in the prior art
And system.
The content of the invention
The object of the present invention is to provide a kind of photo classification method and system, this method and system cause by magnanimity photo with
Individual for unit classification when, it is efficient, at low cost, simple, easy and with very high nicety of grading.
One side according to embodiments of the present invention provides a kind of photo classification system, including:Primary election module, is based on
Multiple base photos are carried out similarity calculation by default similarity algorithm with every reference picture respectively, and by similarity value height
It is stored in an at least base photo for default similarity threshold as primary election photo group to database;Mark sort module,
Based on user to the annotation results of every base photo in the primary election photo group, judge every base photo whether with the base
Quasi- photo is same person;If so, the base photo is copied to the catalogue where the reference picture.
Further, the photo classification system, the mark sort module include:Acquisition request unit is marked, is used for
Obtain the mark request of the triggering generation based on user;Annotation results acquiring unit, for obtaining user in primary election photo group
The annotation results of every base photo;Allocation unit is obtained the primary election not marked from database based on the mark request and shone
Piece group, by user described in the primary election photo component dispensing not marked;Taxon judges every base photo based on annotation results
Whether it is same person with reference picture;If so, in catalogue where the base photo to be categorized into the reference picture.
Further, the photo classification system, the mark sort module, including:Dimension model switch unit, is used for
It is switched over according to the handover request of user between single dimension model and multi-person labeling pattern.
Further, the photo classification system, under multi-person labeling mode state, the mark sort module includes:
Acquisition request unit is marked, obtains the mark request of ordinary user and advanced level user's transmission;Allocation unit, it is general receiving some
After the mark request that general family is sent, judge to whether there is in database and marked primary election photo once by other ordinary users
Group, if in the presence of by ordinary user described in the primary election photo component dispensing;If being not present, obtained not from obtaining in database
The primary election photo group of mark, by ordinary user described in the primary election photo component dispensing not marked;Dispute photo group generation unit, is used
In the similarity for two annotation results for calculating same primary election photo group, and judge two marks of same primary election photo group
Whether similarity as a result is less than predetermined threshold value;If being less than, judge the primary election photo group for photo group of disputing on;The distribution is single
Member is additionally operable to after the mark request of advanced level user's transmission is received, and is judged with the presence or absence of dispute photo group in database, if depositing
Then by the advanced level user of the dispute photo component dispensing request mark;If being not present, obtained not from obtaining in database
The primary election photo group of mark, the primary election photo component dispensing advanced level user that will do not marked.
Further, the photo classification system, further includes:Reference picture replacement module, for obtaining based on user's
The replacement instruction of generation is triggered, and base photo is replaced by the reference picture based on the replacement instruction.
Further, the photo classification system, further includes:Classification results duplicate checking module, for obtaining user to two
The judgement of reference picture based on judgement result as a result, and judge whether two reference pictures are same person;If so, by two
Two catalogues corresponding to reference picture merge into one.
Further, the photo classification system, further includes:Face confines module, for obtaining user to being illustrated in photograph
The operation behavior of the Initial Face frame of on piece, and repaiied based on operation behavior generation coordinate modification information and based on described
Breath modification Initial Face coordinate information is converted to, obtains newer face coordinate information.
Further, the photo classification system, further includes:Feature location module, for obtaining user in request feature
The coordinate information of clicking operation on the photo of positioning;A characteristic point control is generated at clicking operation based on the coordinate information
All characteristic point controls on part and photo based on request feature location generate characteristic point informations.
Other side according to embodiments of the present invention provides a kind of photo classification method, including:Based on default phase
Multiple base photos are subjected to similarity calculation with every reference picture respectively like degree algorithm, and by similarity value higher than default
An at least base photo for similarity threshold is stored as primary election photo group to database;Based on user to the primary election photo
The annotation results of every base photo in group, judge every base photo whether with the reference picture for same person;If so,
The base photo is then copied to the catalogue where the reference picture.
Further, the photo classification method shines every basis in the primary election photo group based on user described
Before the annotation results of piece, further include:Obtain the mark request of the triggering generation based on user;User is obtained to primary election photo group
In every base photo annotation results;The primary election photo group not marked is obtained from database based on the mark request, it will
User described in the primary election photo component dispensing not marked.
Further, the photo classification method, before the mark request of triggering generation of the acquisition based on user,
It further includes:Obtain the handover request of user, and according to the handover request of user single dimension model and multi-person labeling pattern it
Between switch over.
Further, the photo classification method, if the handover request is multi-person labeling mode request, the photo
Sorting technique further includes:After the mark request of some ordinary user transmission is received, judge to whether there is by it in database
His ordinary user marked primary election photo group once, if in the presence of by ordinary user described in the primary election photo component dispensing;If
It is not present, then the primary election photo group not marked is obtained from acquisition database, it will be described in the primary election photo component dispensing that do not marked
Ordinary user;The similarity of two annotation results of same primary election photo group is calculated, and judges same primary election photo group
Whether the similarity of two annotation results is less than predetermined threshold value;If being less than, judge the primary election photo group for photo group of disputing on;
After the mark request for receiving advanced level user's transmission, judge with the presence or absence of dispute photo group in database, if in the presence of by described in
The advanced level user of photo component of disputing on dispensing request mark;If being not present, the primary election not marked is obtained from acquisition database
Photo group, the primary election photo component dispensing advanced level user that will do not marked.
Further, the photo classification method, further includes:Obtain the replacement instruction of the triggering generation based on user;Base
Base photo is replaced into the reference picture in the replacement instruction.
Further, the photo classification method, further includes:Obtain judgement result of the user to two reference pictures;Base
Judge whether two reference pictures are same person in judgement result;If so, by two mesh corresponding to two reference pictures
Record merges into one.
Further, the photo classification method, further includes:User is obtained to being illustrated in the Initial Face frame on photo
Operation behavior;Coordinate modification information is generated based on the operation behavior;Based on modification information modification Initial Face coordinate letter
Breath, obtains newer face coordinate information.
Further, the photo classification method, further includes:It obtains user and behaviour is clicked on the photo of request feature location
The coordinate information of work;A characteristic point control is generated at clicking operation based on the coordinate information;Based on request feature location
Photo on all characteristic point controls generation characteristic point informations.
Another aspect according to embodiments of the present invention provides a kind of electronic equipment for photo classification, including:Extremely
A few processor and, the memory being connected at least one processor communication;Wherein, the memory storage has
The instruction that can be performed by one processor, described instruction are performed by least one processor so that it is described at least
One processor is able to carry out the photo classification method.
The advantageous effect of the embodiment of the present invention is, by the way that the photo of magnanimity is classified in units of individual, obtains
Photo group in units of individual;The photo group includes photo of the same person under different scenes, based on the photo group
Face recognition algorithms, which are trained, can improve the precision of face recognition algorithms.Method and system provided by the invention realizes photograph
Piece employs algorithm and combines the mode manually marked when classifying, compared with the method for being completely dependent on manually classifying, save
Manpower, financial resources;The method classified compared with algorithm is completely dependent on has very high nicety of grading.
Description of the drawings
Fig. 1 is the first embodiment module relationship schematic diagram of photo classification system provided by the invention;
Fig. 2 is a specific example of the first embodiment of photo classification system provided by the invention;
Fig. 3 is the first embodiment flow chart of steps of photo classification method provided by the invention;
Fig. 4 is the hardware architecture diagram of electronic equipment provided by the invention.
Specific embodiment
Understand to make the object, technical solutions and advantages of the present invention clearer, With reference to embodiment and join
According to attached drawing, the present invention is described in more detail.It should be understood that these descriptions are merely illustrative, and it is not intended to limit this hair
Bright scope.In addition, in the following description, the description to known features and technology is omitted, to avoid this is unnecessarily obscured
The concept of invention.
Fig. 1 is the first embodiment module relationship schematic diagram of photo classification system provided by the invention.
As shown in Figure 1, in embodiments of the present invention, photo classification system includes:Primary election module, database and mark classification
Module.
Primary election module carries out multiple base photos for being based on default similarity algorithm with every reference picture respectively
Similarity calculation, and using an at least base photo of the similarity value higher than default similarity threshold as primary election photo group,
And primary election photo group is stored into database.
Wherein, base photo refers to the photo to be sorted of the most original obtained from database.Reference picture refers to
Taking human as the photo that can represent a people in the photo of unit classification.When having multiple pictures in the catalogue of someone, benchmark
Photo can be random one in optimal photo or multiple pictures in multiple pictures.The embodiment of the present invention is not made
It limits.When there was only a photo in the catalogue of someone, which is the reference picture of this people.Wherein, one is judged
Whether photo is that the factor of optimal photo includes clarity, light and shade contrast, brightness, the personage's angle (front, side) of photo
Deng.
Picture similarity calculating method of the prior art may be employed in similarity algorithm.Algorithm is not the present invention in itself
Inventive point, details are not described herein.
Specifically, primary election module includes:Photo acquiring unit, similarity calculated and primary election photo group generation unit.
Wherein, photo acquiring unit, for obtaining a reference picture and multiple base photos;Similarity calculated, for being based on
Reference picture is carried out similarity calculation by default similarity algorithm with multiple base photos respectively so that every base photo obtains
To corresponding similarity value;Primary election photo group generation unit, for being based on default similarity threshold, by similarity value
Higher than default similarity threshold base photo as primary election photo group.
Sort module is marked, based on user to the annotation results of every base photo in the primary election photo group, is judged every
Open whether base photo is same person with the reference picture;If so, the base photo is copied into the reference picture institute
Catalogue in.
Wherein, annotation results include show base photo whether with reference picture be same person field.Using which
It is various that specific character, which is used as field, for example, can define field and be:Yes, No.Wherein, Yes represents base photo
It is same person with reference picture;No represents that base photo and reference picture are not same persons.Field can also be defined:1,0.
Wherein, 1 represent base photo and reference picture for same person;0 represents that base photo and reference picture are not same persons.This
Inventive embodiments are not construed as limiting.
Fig. 2 is a specific example of the first embodiment of photo classification system provided by the invention.
As shown in Fig. 2, in the specific example, base photo is 100,000, previously generates 3 catalogue (photos of Zhang San
Catalogue, the directory of photos of Li Si, the directory of photos of king five), each catalogue include 3 reference pictures (reference picture of Zhang San,
Reference picture, the reference picture of king five of Li Si), default similarity threshold be 60%.
Primary election module be based on default similarity algorithm by 100,000 base photos respectively with the 1st reference picture (Zhang San
Reference picture) carry out similarity calculation, obtain 100,000 similarity values;Afterwards, 100,000 similarity calculation results are distinguished
It is compared with similarity threshold for 60%, only 40 similarity values higher than 60%;Afterwards, by this 40 similarity value institutes
Corresponding base photo stores primary election photo group into database as primary election photo group (the primary election photo of Zhang San).
User is labeled every base photo in the primary election photo of Zhang San, obtains 40 annotation results;
Whether mark sort module judges the corresponding base photo of the annotation results with Zhang San's based on each annotation results
Reference picture is same person;If so, (Zhang San in catalogue where the base photo to be copied to the reference picture of the Zhang San
Directory of photos).So far, the reference picture comparison of Zhang San finishes.
Similar to the comparison process of the reference picture to Zhang San, the reference picture of Li Si is compared successively and the benchmark of king five shines
Piece.After comparison finishes, the comparison of more wheels is carried out to remaining base photo, and after the comparison of Duolun, according to user's
Annotation results establish new person's catalogue respectively for remaining base photo.
In the second embodiment of photo classification system provided by the invention, on the basis of above-mentioned first embodiment, institute
Stating mark sort module includes:Mark acquisition request unit, annotation results acquiring unit, allocation unit and taxon.
Wherein, acquisition request unit is marked, for obtaining the mark request of the triggering generation based on user.The mark please
Ask including:The storage of the primary election photo group of the account information of user, the ID number of the primary election photo group of request mark and request mark
Path.
Annotation results acquiring unit, for obtaining annotation results of the user to every base photo in primary election photo group.
Allocation unit obtains the primary election photo group not marked based on the mark request, by what is do not marked from database
User described in primary election photo component dispensing;
Taxon, based on annotation results judge every base photo whether with reference picture for same person;If so,
The base photo is categorized into the catalogue where the reference picture.
In the 3rd embodiment of photo classification system provided by the invention, in above-mentioned first embodiment or second embodiment
On the basis of, the mark sort module further includes:Dimension model switch unit, for according to the handover request of user in list
It is switched between people's dimension model and multi-person labeling pattern.Wherein, under single dimension model state, the mark classification mould
Block includes every unit described in second embodiment.Under multi-person labeling mode state, the mark sort module includes:Mark
Note acquisition request unit, allocation unit and dispute photo group generation unit.
Wherein, acquisition request unit is marked, obtains the mark request of ordinary user and advanced level user's transmission;Wherein, commonly
The mark request that user sends includes:The account information of ordinary user, the ID number of the primary election photo group of request mark and request mark
The store path of the primary election photo group of note;The mark request that advanced level user sends includes:Account information, the request mark of advanced level user
The store path of the ID number of the primary election photo group of note and the primary election photo group of request mark.
Allocation unit after the mark request of some ordinary user transmission is received, judges to whether there is quilt in database
Other ordinary users marked primary election photo group once, if in the presence of by ordinary user described in the primary election photo component dispensing;
If being not present, the primary election photo group not marked, the primary election photo component dispensing institute that will do not marked are obtained from acquisition database
State ordinary user.
Dispute photo group generation unit, for calculating the similarity of two annotation results of same primary election photo group, and
Judge whether the similarity of two annotation results of same primary election photo group is less than predetermined threshold value;If being less than, judge that this is first
Photo group is selected as dispute photo group.Wherein, two annotation results of same primary election photo group are by two different common use
What family marked respectively.
Calculating the method for the similarity of two annotation results of same primary election photo group is:Not phase in two annotation results
The same quantity of field divided by the number of pictures of primary election photo group.Such as:There are 10 photos in primary election photo group A.Ordinary user
First is successively to the annotation results of 10 photos:Yes、Yes、Yes、Yes、Yes、Yes、No、No、No、No.Ordinary user's second pair
The annotation results of 10 photos are successively:Yes、No、No、No、No、No、No、No、No、No.As can be seen that two ordinary users
Annotation results in have 5 differences, then the similarity of two annotation results be 5/10=50%.
The allocation unit is additionally operable to after the mark request of advanced level user's transmission is received, judge in database whether
In the presence of dispute photo group, if in the presence of by the advanced level user of the dispute photo component dispensing request mark;If being not present,
The primary election photo group not marked, the primary election photo component dispensing advanced level user that will do not marked are obtained from acquisition database.Wherein,
Photo group dispute on to have there are two annotation results, and the similarity of two annotation results is less than the primary election photo group of predetermined threshold value.
By taking above-mentioned example as an example, if predetermined threshold value is 80%, primary election photo group A is dispute photo group.
The further of above-mentioned 3rd embodiment has the beneficial effect that:The pattern of multi-person labeling is provided, can effectively be subtracted
Few manually-operated error.
In the fourth embodiment of photo classification system provided by the invention, on the basis of above-mentioned first embodiment, shine
Piece categorizing system further includes:Reference picture replacement module for obtaining the replacement instruction of the triggering generation based on user, and is based on
Base photo is replaced the reference picture by the replacement instruction.
Wherein, in front end, user, which triggers, replaces button, generates replacement instruction.Wherein, replacement instruction includes:User is signified
The ID number of fixed base photo, the store path of the base photo, the ID number of the reference picture to be replaced, the base to be replaced
The store path of quasi- photo.
Base photo is replaced reference picture based on replacement instruction includes:It is assigned newly for the reference picture to be replaced
ID number;The store path of the base photo is changed to the store path of the reference picture to be replaced;It is shone for the basis
Piece assigns the ID number of former reference picture.
In the 5th embodiment of photo classification system provided by the invention, on the basis of above-mentioned first embodiment, shine
Piece categorizing system further includes classification results duplicate checking module, for obtaining user to the judgement of two reference pictures as a result, and being based on
Judge that result judges whether two reference pictures are same person;If so, by two catalogues corresponding to two reference pictures
Merge into one.
Specifically, classification results duplicate checking module includes:Duplicate checking acquisition request unit, reference picture acquiring unit, similarity
Computing unit, reference picture allocation unit judge result acquiring unit and combining unit.
Wherein, duplicate checking acquisition request unit obtains the duplicate checking request of the triggering generation based on user.Reference picture obtains single
Member obtains the reference picture in each catalogue based on duplicate checking request from database.Similarity calculated, by certain
One reference picture carries out similarity calculation with other reference pictures respectively and obtains similarity value.Reference picture allocation unit, will
Highest two reference pictures of similarity value distribute to the user of request duplicate checking.Result acquiring unit is judged, for obtaining user
To the judgement result of two reference pictures.Combining unit judges whether two reference pictures are same person based on judgement result;
If so, two catalogues corresponding to two reference pictures are merged into one.
In the sixth embodiment of photo classification system provided by the invention, on the basis of above-mentioned first embodiment, shine
Piece categorizing system further includes:Face confines module, for obtaining user to being illustrated in the operation row of the Initial Face frame on photo
For, and based on operation behavior generation coordinate modification information and based on modification information modification Initial Face coordinate letter
Breath, obtains newer face coordinate information.
Specifically, face is confined module and is included:Confine acquisition request unit, coordinate calculating unit, the generation of Initial Face frame
Unit, display unit, user behavior acquiring unit, coordinate modification information generation unit and face coordinate information generation unit.
Wherein, acquisition request unit is confined, obtain the triggering generation based on user confines request;It is described to confine request bag
It includes:The store path for the photo that the ID number for the photo that the account information of user, request are confined and request are confined.
Coordinate calculating unit, the photo confined based on default algorithm to request are calculated, and obtain Initial Face coordinate
Information.Initial Face frame generation unit generates Initial Face frame based on Initial Face coordinate information.Display unit, in front end
The Initial Face frame is shown on interface.User behavior acquiring unit obtains operation behavior of the user to Initial Face frame.It is described
Operation behavior includes:Drag the position of Initial Face frame, the size of change Initial Face frame, deletion Initial Face frame.Coordinate is repaiied
Change information generating unit, coordinate modification information is generated based on the operation behavior.Face coordinate information generation unit, based on described
Modification information changes the Initial Face coordinate information, obtains newer face coordinate information, and newer face coordinate is believed
Breath is stored to the catalogue where the photo.Wherein, newer face coordinate information is the TXT file of the same name with photo.
The further of the embodiment of the present invention has the beneficial effect that:Provide a kind of change face coordinate letter of man-machine interactive
The mode of breath determines face coordinate information so that face coordinate information is more accurate, this is sent out compared with essence is deposited by algorithm
The face coordinate information of bright embodiment generation for the training to face recognition algorithms, can further effective boosting algorithm standard
True rate.
In the sixth embodiment of photo classification system provided by the invention, on the basis of above-mentioned first embodiment, shine
Piece categorizing system further includes feature location module, for obtaining the coordinate of user's clicking operation on the photo of request feature location
Information;A characteristic point control and the photograph based on request feature location are generated at clicking operation based on the coordinate information
All characteristic point control generation characteristic point informations of on piece.
Specifically, feature location module includes:Location Request acquiring unit, feature templates display unit, characteristic point control
Generation unit and characteristic point information generation unit.
Wherein, Location Request acquiring unit obtains the feature location request of the triggering generation based on user.The feature is determined
Position request includes:The storage of the photo of the account information of user, the ID number of the photo of request feature location and request feature location
Path.
Feature templates display unit, for showing feature templates.The feature templates include template facial image and display
In multiple characteristic points on template facial image.Position of multiple characteristic points on the template facial image is:Positioned at eyebrow
Left end, right end and centre position.Positioned at eyes left end, right end and centre position.Positioned at nose left end, right end and centre position.
Positioned at face left end, right end and centre position.Positioned at chin lower end.Positioned at ear lower end.
Characteristic point control generation unit obtains the coordinate information of user's clicking operation on the photo of request feature location.
A characteristic point control is generated at clicking operation based on the coordinate information.The characteristic point control is monitoring the point of user
When hitting operation, color is become into the second color from the first color, and never editable pattern is changed to editable pattern by pattern.
The characteristic point control is moved under editable pattern when monitoring the moving operation of user.The characteristic point control
Under editable pattern, deleted when monitoring the delete operation of user.
Characteristic point information generation unit generates features based on all characteristic point controls on the photo for asking feature location
Point information, and characteristic point information is stored to the catalogue where the photo of request feature location.Wherein, characteristic point information is with asking
Seek the TXT files that the photo of feature location is of the same name.The characteristic point information includes:It asks all on the photo of feature location
The coordinate information of characteristic point control.
The further of the embodiment of the present invention has the beneficial effect that:Provide a kind of definite human face characteristic point of man-machine interactive
Mode, by algorithm determine human face characteristic point compared with essence is deposited so that human face characteristic point is more accurate, and the present invention is implemented
The human face characteristic point of example generation for training to face recognition algorithms, can further effective boosting algorithm accuracy rate.
Fig. 3 is the first embodiment flow chart of steps of photo classification method provided by the invention.
As shown in figure 3, in the present embodiment, photo classification method comprises the following steps:
Multiple base photos are carried out similarity meter by S1 with every reference picture respectively based on default similarity algorithm
It calculates, obtains similarity value;
S2 stores an at least base photo of the similarity value higher than default similarity threshold as primary election photo group
To database;
S3 based on user to the annotation results of every base photo in the primary election photo group, judges every base photo
Whether it is same person with the reference picture;If so, perform S4.
S4 copies to the base photo in the catalogue where the reference picture.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, wherein, S3 based on user to the primary election photo group in every base photo annotation results before, further include:
Obtain the mark request of the triggering generation based on user;
Obtain annotation results of the user to every base photo in primary election photo group;
The primary election photo group not marked, the primary election photo group that will do not marked are obtained from database based on the mark request
Distribute to the user.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, wherein, before the mark request of triggering generation of the acquisition based on user, further include:Obtain the switching of user
Request, and switched over according to the handover request of user between single dimension model and multi-person labeling pattern.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, wherein, if the handover request is multi-person labeling mode request, the photo classification method further includes:
After the mark request of some ordinary user transmission is received, judge to whether there is in database by other common use
Family marked primary election photo group once, if in the presence of by ordinary user described in the primary election photo component dispensing;If being not present,
The primary election photo group not marked is then obtained from acquisition database, will commonly be used described in the primary election photo component dispensing not marked
Family;
The similarity of two annotation results of same primary election photo group is calculated, and judges the two of same primary election photo group
Whether the similarity of a annotation results is less than predetermined threshold value;If being less than, judge the primary election photo group for photo group of disputing on;
After the mark request of advanced level user's transmission is received, judge with the presence or absence of dispute photo group in database, if depositing
Then by the advanced level user of the dispute photo component dispensing request mark;If being not present, obtained not from obtaining in database
The primary election photo group of mark, the primary election photo component dispensing advanced level user that will do not marked.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, it further includes:Obtain the replacement instruction of the triggering generation based on user;Base photo is replaced based on the replacement instruction
Fall the reference picture.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, it further includes:Obtain judgement result of the user to two reference pictures;Judge that two reference pictures are based on judgement result
No is same person;If so, two catalogues corresponding to two reference pictures are merged into one.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, it further includes:User is obtained to being illustrated in the operation behavior of the Initial Face frame on photo;It is given birth to based on the operation behavior
Into coordinate modification information;Initial Face coordinate information is changed based on the modification information, obtains newer face coordinate information.
In another embodiment of photo classification method provided by the invention, in the embodiment of above-mentioned photo classification method
On the basis of, it further includes:Obtain the coordinate information of user's clicking operation on the photo of request feature location;Believed based on the coordinate
Breath generates a characteristic point control at clicking operation;All characteristic point control lifes on photo based on request feature location
Into characteristic point information.
Fig. 4 is the hardware architecture diagram provided by the present invention for the electronic equipment of photo classification.
As shown in figure 4, the present invention also provides a kind of electronic equipment for photo classification, including:At one or more
Device and memory are managed, in Fig. 4 by taking a processor as an example.
It can also include for the electronic equipment of photo classification:Input unit and output device.
Processor, memory, input unit and output device can be connected by bus or other modes, with logical in Fig. 4
It crosses exemplified by the mode of bus connection.
The present invention is implemented it will be understood by those skilled in the art that the structure of the electronic equipment shown in Fig. 4 is not formed
The restriction of example, it can also include either busbar network or hub-and-spoke configuration than illustrating more or fewer portions
Part either combines some components or different components arrangement.
Processor can be made of integrated circuit (IntegratedCircuit, abbreviation IC), such as can be by single encapsulation
IC formed, can also be formed by connecting the encapsulation IC of more identical functions or difference in functionality.For example, processor can
Only to include central processing unit (CentralProcessingUnit, abbreviation CPU) or CPU, digital signal processor
(digitalsignalprocessor, abbreviation DSP), graphics processor (GraphicProcessingUnit, abbreviation GPU) and
The combination of various control chips.In embodiments of the present invention, CPU can be single operation core, can also include multioperation core
The heart.
Memory as a kind of non-transient computer readable storage medium storing program for executing, available for store non-transient software program, it is non-temporarily
State computer class performs program and module, such as the corresponding program instruction/module of photo classification system in the embodiment of the present application
(for example, attached primary election module shown in FIG. 1 and mark sort module).Processor is stored in the non-transient soft of memory by operation
Part program, instruction and module, various function application and data processing so as to execute server, that is, realize above-mentioned photo point
The processing method of class embodiment of the method.
Memory can include storing program area and storage data field, wherein, storing program area can storage program area, extremely
A few required application program of function;In embodiments of the present invention, operating system can be android system, iOS system
Or Windows operating system etc..Storage data field can be stored uses created data etc. according to photo classification system.This
Outside, memory can include high-speed random access memory, can also include non-transient memory, and a for example, at least disk is deposited
Memory device, flush memory device or other non-transient solid-state memories.In some embodiments, memory is optional including opposite
In the remotely located memory of processor.The embodiment of above-mentioned network includes but not limited to internet, intranet, local
Net, mobile radio communication and combinations thereof.
Input unit can receive the number of input or the user of character information and the processing unit of generation and list operation
Set and function control it is related key signals input.Input unit can include touch-screen, keyboard, mouse etc., can also wrap
Include wireline interface, wireless interface etc..
Output device may include the equipment such as display screen, loud speaker, can also include wireline interface, wireless interface etc..
Electronic equipment can be smart mobile phone (such as Android phone, iOS mobile phones), tablet computer, palm PC and
Mobile internet device MobileInternetDevices, MID), PAD etc..
It should be appreciated that the above-mentioned specific embodiment of the present invention is used only for exemplary illustration or explains the present invention's
Principle, without being construed as limiting the invention.Therefore, that is done without departing from the spirit and scope of the present invention is any
Modification, equivalent substitution, improvement etc., should all be included in the protection scope of the present invention.In addition, appended claims purport of the present invention
Covering the whole variations fallen into scope and border or this scope and the equivalents on border and repairing
Change example.
Claims (10)
1. a kind of photo classification system, which is characterized in that including:
Multiple base photos are carried out similarity meter by primary election module with every reference picture respectively based on default similarity algorithm
It calculates, and an at least base photo of the similarity value higher than default similarity threshold is stored as primary election photo group to data
Storehouse;
Sort module is marked, based on user to the annotation results of every base photo in the primary election photo group, judges every base
Whether plinth photo is same person with the reference picture;If so, the base photo is copied to where the reference picture
Catalogue.
2. photo classification system according to claim 1, which is characterized in that the mark sort module includes:
Acquisition request unit is marked, for obtaining the mark request of the triggering generation based on user;
Annotation results acquiring unit, for obtaining annotation results of the user to every base photo in primary election photo group;
Allocation unit obtains the primary election photo group not marked, the primary election that will do not marked based on the mark request from database
User described in photo component dispensing;
Taxon, based on annotation results judge every base photo whether with reference picture for same person;It if so, should
Base photo is categorized into the catalogue where the reference picture.
3. photo classification system according to claim 1, which is characterized in that wherein, the mark sort module, including:
Dimension model switch unit is cut for the handover request according to user between single dimension model and multi-person labeling pattern
It changes.
4. photo classification system according to claim 3, which is characterized in that under multi-person labeling mode state, the mark
Note sort module includes:
Acquisition request unit is marked, obtains the mark request of ordinary user and advanced level user's transmission;
Allocation unit after the mark request of some ordinary user transmission is received, judges to whether there is by other in database
Ordinary user marked primary election photo group once, if in the presence of by ordinary user described in the primary election photo component dispensing;If no
In the presence of, then the primary election photo group not marked is obtained from acquisition database, it will be general described in the primary election photo component dispensing not marked
General family;
Dispute photo group generation unit, for calculating the similarity of two annotation results of same primary election photo group, and judges
Whether the similarity of two annotation results of same primary election photo group is less than predetermined threshold value;If being less than, judge that the primary election is shone
Piece group is dispute photo group;
The allocation unit is additionally operable to after the mark request of advanced level user's transmission is received, judges to whether there is in database
Dispute photo group, if in the presence of by the advanced level user of the dispute photo component dispensing request mark;If being not present, from obtaining
The primary election photo group for obtaining in database and not marking is taken, the primary election photo component dispensing advanced level user that will do not marked.
5. photo classification system according to claim 1, which is characterized in that further include:Reference picture replacement module, is used for
The replacement instruction of the triggering generation based on user is obtained, and base photo is replaced by the benchmark based on the replacement instruction and is shone
Piece.
6. photo classification system according to claim 1, which is characterized in that further include:Classification results duplicate checking module, is used for
User is obtained to the judgement of two reference pictures as a result, and judging whether two reference pictures are same based on judgement result
People;If so, two catalogues corresponding to two reference pictures are merged into one.
7. photo classification system according to claim 1, which is characterized in that further include:Face confines module, for obtaining
User generates coordinate modification information to being illustrated in the operation behavior of the Initial Face frame on photo based on the operation behavior,
And Initial Face coordinate information is changed based on the modification information, obtain newer face coordinate information.
8. photo classification system according to claim 1, which is characterized in that further include:Feature location module, for obtaining
The coordinate information of user's clicking operation on the photo of request feature location;It is generated based on the coordinate information at clicking operation
All characteristic point controls on one characteristic point control and photo based on request feature location generate characteristic point informations.
9. a kind of photo classification method, which is characterized in that including:
Multiple base photos are carried out by similarity calculation with every reference picture based on default similarity algorithm respectively, it will be similar
Angle value is stored as primary election photo group to database higher than an at least base photo for default similarity threshold;
Based on user to the annotation results of every base photo in the primary election photo group, judge every base photo whether with institute
Reference picture is stated as same person;If so, in catalogue where the base photo to be copied to the reference picture.
10. a kind of electronic equipment for photo classification, including:At least one processor and, at least one processing
The memory of device communication connection;Wherein,
The memory storage has the instruction that can be performed by one processor, and described instruction is by least one processor
It performs, so that at least one processor is able to carry out the photo classification method described in claim 9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711231374.4A CN108052555A (en) | 2017-11-29 | 2017-11-29 | A kind of photo classification method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711231374.4A CN108052555A (en) | 2017-11-29 | 2017-11-29 | A kind of photo classification method and system |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108052555A true CN108052555A (en) | 2018-05-18 |
Family
ID=62121249
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711231374.4A Pending CN108052555A (en) | 2017-11-29 | 2017-11-29 | A kind of photo classification method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108052555A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108960297A (en) * | 2018-06-15 | 2018-12-07 | 北京金山云网络技术有限公司 | Mask method, annotation equipment, equipment and the storage medium of picture |
CN112765384A (en) * | 2020-12-31 | 2021-05-07 | 北京华安信联通信技术有限公司 | Photo classification method |
-
2017
- 2017-11-29 CN CN201711231374.4A patent/CN108052555A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108960297A (en) * | 2018-06-15 | 2018-12-07 | 北京金山云网络技术有限公司 | Mask method, annotation equipment, equipment and the storage medium of picture |
CN108960297B (en) * | 2018-06-15 | 2021-07-30 | 北京金山云网络技术有限公司 | Picture labeling method, labeling device, equipment and storage medium |
CN112765384A (en) * | 2020-12-31 | 2021-05-07 | 北京华安信联通信技术有限公司 | Photo classification method |
CN112765384B (en) * | 2020-12-31 | 2024-01-16 | 北京华安信联通信技术有限公司 | Photo classification method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11205023B2 (en) | Computer aided systems and methods for creating custom products | |
US10956784B2 (en) | Neural network-based image manipulation | |
CN108197532B (en) | The method, apparatus and computer installation of recognition of face | |
US10049308B1 (en) | Synthesizing training data | |
CN102236890B (en) | From multiple Computer image genration combination image | |
CN105354248B (en) | The recognition methods of distributed image low-level image feature and system based on gray scale | |
CN110135246A (en) | A kind of recognition methods and equipment of human action | |
CN106874826A (en) | Face key point-tracking method and device | |
JP2017504861A (en) | Image-based search | |
CN110414428A (en) | A method of generating face character information identification model | |
KR102045575B1 (en) | Smart mirror display device | |
US20140293069A1 (en) | Real-time image classification and automated image content curation | |
CN104899825A (en) | Method and device for modeling picture figure | |
CN106233228A (en) | Process the method for content and use the electronic equipment of the method | |
CN108229367A (en) | A kind of face identification method and device | |
WO2007129474A1 (en) | Object recognition device, object recognition program, and image search service providing method | |
Saboia et al. | Eye specular highlights telltales for digital forensics: A machine learning approach | |
CN109800731A (en) | Fingerprint input method and relevant apparatus | |
US11574392B2 (en) | Automatically merging people and objects from multiple digital images to generate a composite digital image | |
CN110222573A (en) | Face identification method, device, computer equipment and storage medium | |
CN108647351A (en) | Processing method, device, storage medium and the terminal of text image | |
CN107516105A (en) | Image processing method and device | |
TW202009682A (en) | Interactive method and device based on augmented reality | |
CN109886153A (en) | A kind of real-time face detection method based on depth convolutional neural networks | |
CN109598748B (en) | Image extraction device, image extraction method, image extraction program, and recording medium storing the program |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180518 |
|
WD01 | Invention patent application deemed withdrawn after publication |