CN107451194A - A kind of image searching method and device - Google Patents
A kind of image searching method and device Download PDFInfo
- Publication number
- CN107451194A CN107451194A CN201710527201.0A CN201710527201A CN107451194A CN 107451194 A CN107451194 A CN 107451194A CN 201710527201 A CN201710527201 A CN 201710527201A CN 107451194 A CN107451194 A CN 107451194A
- Authority
- CN
- China
- Prior art keywords
- atlas
- picture
- search
- search result
- submodule
- 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
Abstract
The application provides a kind of image searching method and device, including：Receive search term；Search result items are obtained from picture database and atlas database according to the search term；Hybrid-sorting is carried out to the search result items；Show the search result items after the sequence.It can avoid not using search engine technique to build storehouse and retrieval ordering to atlas in the prior art, it is necessary to excavate atlas resource, scalability and ageing all poor for high frequency search in advance；The feature for not introducing atlas dimension participates in sequence, it is impossible to the effectively correlation and quality of control atlas；Do not introduce click feature and form negative feedback mechanism, low-quality atlas can not be in system the problem of natural subsidence.User can be reduced the cost of figure is selected in search result, while meet the needs of user is to complete picture.
Description
【Technical field】
The application is related to the Internet, applications field, more particularly to a kind of image searching method and device.
【Background technology】
Picture retrieval (Image Search) refers to that user inputs natural language, is searched from picture set and presses correlation
Etc. index, information retrieval (Information Retrieval) process of image results to user of sequence is returned.
Photographic search engine (Image Search Engine) is exactly the information retrieval for being used to search Internet picture information
Instrument.Existing photographic search engine is that single picture is recalled and sorted, and search result is deployed by picture.This scheme expires
The demand that foot user on the internet " looks for figure ", but the presentation mode experience of result page is bad.At pc ends, result page often shields can
So that the thumbnail result of 10-20 pictures is presented；But in mobile terminal, result page, which often shields, can only be presented 4-6 pictures results.
Existing photographic search engine is not retrieved for atlas, the atlas result under only a small amount of high frequency search, institute
The scheme of use be previously according to Topics Crawling atlas resource and before be inserted into search result.Such scheme has the following disadvantages：
1) search engine technique is not used to build storehouse and retrieval ordering to atlas, it is necessary to be excavated in advance for high frequency search
Atlas resource, scalability and ageing all poor；
2) sequence is participated in without the feature for introducing atlas dimension, it is impossible to the effectively correlation and quality of control atlas；
3) without introduce click feature formed negative feedback mechanism, low-quality atlas can not in system natural subsidence.
【The content of the invention】
The many aspects of the application provide a kind of image searching method and device, to provide atlas search result.
The one side of the application, there is provided a kind of image searching method, including：
Receive search term；
Search result items are obtained from picture database and atlas database according to the search term；
Hybrid-sorting is carried out to the search result items；
Show the search result items after the sequence.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described according to institute
Stating search term search result items are obtained from picture database and atlas database includes following sub-step：
Scanned in the picture inverted index and atlas inverted index pre-established, acquisition matches with the search term
Index；
Corresponding with the index of the search term matching picture and atlas are obtained, generates search result items.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, described in reception
Before search term, network picture is captured, establishes inverted index, including following sub-step：
The structured text field of webpage where the picture of crawl is analyzed, obtains the text message of the picture；
The picture of crawl is excavated, generates atlas；
Establish atlas inverted index.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, described pair of crawl
Picture excavated, generation atlas include：
The picture of crawl is polymerize, obtains intelligence polymerization atlas.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, described pair of crawl
Picture excavated, generation atlas include：
By the constitutive characteristic of network address is similar and descriptor identical picture generation webpage atlas
Aspect as described above and any possible implementation, it is further provided a kind of implementation, it is described to described
Search result items carry out hybrid-sorting and further comprise following sub-step：
Feature extraction is carried out to picture and atlas；
The characteristic of term and picture to be sorted and atlas is inputted to the order models of training in advance, to the respectively row for the treatment of
Sequence picture and atlas carry out hybrid-sorting.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the displaying institute
Stating the search result items after sequence includes：
The thumbnail of picture and atlas is illustrated on result of page searching by hybrid-sorting order.
Another aspect of the present invention, there is provided a kind of atlas device, including：
Receiving module, for receiving search term；
Search module, for obtaining search result items from picture database and atlas database according to the search term；
Order module, for carrying out hybrid-sorting to the search result items；
Display module, for showing the search result items after the sequence.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the search mould
Block includes following submodule：
Acquisition submodule is indexed, for being scanned in the picture inverted index and atlas inverted index pre-established,
Obtain the index matched with the search term；
Search result items generate submodule, for obtaining corresponding with the index of the search term matching picture and atlas,
Generate search result items.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the search mould
Block also includes inverted index setting up submodule, for before the search term is received, being captured to network picture, establishes figure
Collect inverted index, the inverted index setting up submodule includes：
Text message acquisition submodule, the structured text field for webpage where the picture to crawl are analyzed,
Obtain the text message of the picture；
Atlas generates submodule, for being excavated to the picture of crawl, generates atlas；
Atlas inverted index setting up submodule, for establishing atlas inverted index.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the atlas life
Specifically performed into submodule：
The picture of crawl is polymerize, obtains intelligence polymerization atlas.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the atlas life
Specifically performed into submodule：
By the constitutive characteristic of network address is similar and descriptor identical picture generation webpage atlas
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the sequence mould
Block includes following submodule：
Feature extraction submodule, for carrying out feature extraction to picture and atlas；
The hybrid-sorting submodule, it is advance for the characteristic of term and picture to be sorted and atlas to be inputted
The order models of training, hybrid-sorting is carried out to respectively picture and atlas to be sorted.
Aspect as described above and any possible implementation, it is further provided a kind of implementation, the displaying mould
Block specifically performs：
The thumbnail of picture and atlas is illustrated on result of page searching by hybrid-sorting order.
The another aspect of the application, there is provided a kind of equipment, it is characterised in that the equipment includes：
One or more processors；
Storage device, for storing one or more programs,
When one or more of programs are by one or more of computing devices so that one or more of processing
Device realizes any above-mentioned method.
The another aspect of the application, there is provided a kind of computer-readable recording medium, be stored thereon with computer program, it is special
Sign is that the program realizes any above-mentioned method when being executed by processor.
From the technical scheme, the embodiment of the present application obtains the pictorial information of camera shooting, and display is drawn
The pictorial information of AR information is stated, user can be helped to carry out another or the multiple terminals to be looked for of fast positioning.
【Brief description of the drawings】
, below will be to embodiment or description of the prior art in order to illustrate more clearly of the technical scheme in the embodiment of the present application
In the required accompanying drawing used be briefly described, it should be apparent that, drawings in the following description are some realities of the application
Example is applied, for those of ordinary skill in the art, without having to pay creative labor, can also be attached according to these
Figure obtains other accompanying drawings.
Fig. 1 is the schematic flow sheet for the image searching method that the embodiment of the application one provides；
Fig. 2 is from picture database and figure in the image searching method that the embodiment of the application one provides according to the search term
Collect the schematic flow sheet that search result items are obtained in database；
Fig. 3 is before receiving the search term in the image searching method that the embodiment of the application one provides, to network picture
Captured, establish the schematic flow sheet of inverted index
Fig. 4 is to carry out hybrid-sorting to the search result items in the image searching method that the embodiment of the application one provides
Schematic flow sheet；
Fig. 5 is the structural representation for the picture searching device that another embodiment of the application provides；
Fig. 6 is the structural representation of the search module for the picture searching device that another embodiment of the application provides；
Fig. 7 is that the picture inverted index for the picture searching device that another embodiment of the application provides and atlas inverted index are built
Vertical sub-modular structure schematic diagram；
Fig. 8 is the structural representation of the order module for the picture searching device that another embodiment of the application provides；
Fig. 9 is suitable for for realizing the block diagram of the exemplary computer system/server of the embodiment of the present invention.
【Embodiment】
To make the purpose, technical scheme and advantage of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In accompanying drawing, the technical scheme in the embodiment of the present application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, rather than whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art
The whole other embodiments obtained under the premise of creative work is not made, belong to the scope of the application protection.
In addition, the terms "and/or", only a kind of incidence relation for describing affiliated partner, represents there may be
Three kinds of relations, for example, A and/or B, can be represented：Individualism A, while A and B be present, these three situations of individualism B.Separately
Outside, character "/" herein, it is a kind of relation of "or" to typically represent forward-backward correlation object.
Fig. 1 is the schematic flow sheet for the image searching method that the embodiment of the application one provides, as shown in figure 1, including following
Step：
In 101, search term is received；
The search term is encapsulated in Fa send Give search engines in searching request by browser, for asking the search engine
The search picture and atlas related to the search term.
In 102, search result items are obtained from picture database and atlas database according to the search term；Specifically
, as shown in Fig. 2 including following sub-step：
In 201, before the photographic search engine receives the search term, in the picture inverted index pre-established and
Scanned in atlas inverted index, obtain the index matched with the search term；
Preferably, scanned for respectively in the picture inverted index and atlas inverted index pre-established, acquisition and institute
State the index of search term matching.
In 202, corresponding with the index of the search term matching picture and atlas are obtained, generates search result items.
Preferably, picture inverted index and atlas inverted index are established respectively.
Wherein, it is highly developed to establish picture inverted index, be not discussed in detail.
Further, before receiving the search term, network picture is captured, establishes inverted index, such as Fig. 3 institutes
Show, including following sub-step：
In 301, the structured text field of webpage where the picture of crawl is analyzed, obtains the master of the picture
Epigraph；The structured text field of the place webpage of the picture includes the Web page subject description field of the webpage, picture week
Side the text field.
Specifically, cutting word processing is carried out to the Web page subject description field of webpage where the picture, by cutting word processing
As a result the input as key phrases extraction.The descriptor extracted herein, also can be more in addition to usual used word
The compound word of accurate more polynary expression picture semantic, i.e., be made up of two or more collocations.
It is that the result that the text message of above-mentioned picture is carried out to cutting word processing is carried out when wherein extracting word as descriptor
Stop words filters, and extracts the word of default part of speech as descriptor, is typically to extract proper noun as descriptor.
It is to be carried out from the text message of above-mentioned picture at cutting word when extracting the collocation of more than two words as descriptor
Extraction meets the collocation of the two or more word of default Collocation pattern as descriptor in the result of reason.
In 302, the picture of crawl is excavated, generates atlas；
Webpage atlas and intelligence polymerization atlas are divided into according to excavation mode difference, by resulting each webpage atlas, intelligence
Atlas can be polymerize and be combined into atlas resource set.
Include the subject description of atlas, the description of the binary content and picture of picture on the content composition of the atlas
Text (i.e. periphery text of the picture on webpage).
In a kind of implementation of the present embodiment, the picture of crawl is polymerize, obtains intelligence polymerization atlas.
Preferably, it is polymerize according to its descriptor.
Preferably, it is polymerize for example, by the picture correlation technique of the various prior arts of mode identification technology.
In another implementation of the present embodiment, same webpage or continuous webpage are given birth to by the good picture of subject editing
Into webpage atlas, specifically, including following sub-step：
A secondary picture is randomly selected from database as the first picture, the search second picture similar to the first picture.
The database can be stored with where searched in advance engine from the picture and picture that network (for example, internet) is collected or is captured
Webpage.The picture for being more than predetermined threshold with the similarity of the first picture, or descriptor identical picture can be searched for from database
As second picture；For example, it can be searched for for example, by the picture correlation technique of the various prior arts of mode identification technology
Second picture.
Webpage where obtaining each second image from image data base.It should be understood that due to second image has can
It can be present in multiple webpages, therefore, for each second image, at least one webpage can be obtained.
For each webpage, the 3rd image that the link of at least level deep of the webpage is pointed to is obtained.For example, webpage
Link on the chained representation of the first order depth webpage, the net that the link on the second level chained representation of the webpage webpage is pointed to
Link on page, by that analogy.
Preferably, the image in the webpage is obtained as the 3rd image.In addition, when at least level deep of the webpage
Link in exist represent page turning link when, obtain represent page turning link pointed by webpage in image as the 3rd figure
Picture.It can determine that the link indicates whether page turning by descriptive text (for example, page up, lower one page) of link etc..
Area (that is, resolution ratio) is selected to be more than the image of predetermined threshold as the 4th image among the 3rd image.So,
Unessential small figure, corner figure etc. can be filtered.
According to the constitutive characteristic and its descriptor of the network address of the 4th image, the 4th image is grouped, to obtain at least
One webpage atlas.
Generally, the most contents of the network address of complete image are identicals, are only that the difference of numbering (for example, network address
Content before last level separator "/" is identical, and content afterwards is different, i.e., described picture be located at unified webpage or
Continuous webpage), thus can the constitutive characteristic of network address is the similar and image of descriptor identical the 4th be divided into one group, as one
Webpage atlas.
In 303, atlas inverted index is established；
Preferably, the knot that text can will be described to the Web page subject of webpage where the picture of each atlas do cutting word processing
Descriptor of the fruit as atlas；Establish atlas inverted index.
Preferably, can be using the picture in the picture of the crawl in addition to atlas resource set as non-atlas picture.Establish
The picture inverted index of the non-atlas picture.
In 103, hybrid-sorting is carried out to the search result items.
The search result items include picture and atlas and its URL corresponding to the index of search term matching；To described
Search result items carry out hybrid-sorting and carry out hybrid-sorting to picture and atlas；Specifically, as shown in figure 4, including following son
Step：
In 401, feature extraction is carried out to picture and atlas, including：Picture and atlas text (picture periphery text/figure
Collect subject description) data, content-data, qualitative character data, click feature data.
During the hybrid-sorting of atlas and picture, it is preferred that atlas feature is alignd with picture feature, if there is
Corresponding atlas feature, uses the feature of atlas；If there is no corresponding atlas feature, then using the feature of atlas head figures
Alignment；Picture and atlas are ranked up according to the feature after the alignment of picture and atlas.
In 402, the characteristic of term and picture to be sorted and atlas is inputted to the order models of training in advance,
Treat sequence picture and atlas carries out hybrid-sorting.
Specifically, the characteristic of term and picture to be sorted and atlas is inputted to the order models of training in advance,
The respectively relativity measurement value between picture and atlas to be sorted and the term；, will based on the relativity measurement value
Respectively picture and atlas to be sorted carry out hybrid-sorting.
The order models are deep-neural-network, and the original deep-neural-network built in advance is entered using training sample
Row training obtains.
The original deep-neural-network includes：Represent vector generation network and correlation computations network, it is described represent to
Amount generation network is used to different types of data in the training sample being converted to expression vector and inputted to the correlometer
Network is calculated, the correlation computations network is used to the expression vector of input being converted to a relativity measurement value；The training sample
This includes term and picture and atlas characteristic.
In a preferred embodiment of the present embodiment,
The correlation computations network can include：Hidden layer collection and the output being connected with the output end of the hidden layer collection
Layer；Wherein, the hidden layer collection includes one or more end to end hidden layer, the expression of the vector generation network to
Amount output end is connected with the input of the hidden layer collection, and the output layer exports the relativity measurement value.
Because picture and atlas characteristic include：Atlas text (picture periphery text/atlas subject description) data, figure
Piece content-data, qualitative character data, click feature data.It is corresponding, it is described to represent to include five in vector generation network
Vectorial generation unit is represented, is respectively used to associate in the term of input, atlas text data, image content data and picture
Characteristic is converted to corresponding expression vector, is worked with carrying out follow-up model training.
Wherein, it is described to represent vectorial generation unit, a variety of implementations can be had according to the difference of task object：
The expression vector generation of image content data at present using it is wide be CNN (Convolutional Neural
Network, convolutional neural networks) sorter network, the input of the network is the normalized picture pixels matrix of size, is exported as figure
The class probability distribution of piece represents vector, and classification expression vector is usually picture in a picture category complicated variant system (figure classification system
The general class label for having thousand grades to ten thousand grades) on class probability distribution vector A1,A2,…,An.Wherein, A=(A1,A2,…,
An)；Ai(i=1,2 ..., n) is that the picture that CNN networks provide belongs to the probability of i-th of classification, and n is the size (class of classification system
Other number).
Because atlas text data and term are text, therefore the two represents vectorial generating mode phase one
Cause, be the expression vector generation of text.
Text first passes through participle, then each participle according to default dictionary be mapped as an one-hot (solely heat) characterize to
Amount.Such as：(..., 0 ..., 1 ..., 0 ...), the vector length is the size of dictionary, and it is 1 to have an element, and remaining element is whole
For 0, the position number where element 1 corresponds to sequence number of the word in dictionary.Ensuing processing can have several selections, example
Such as BoW-DNN (Bag of Words-Deep Neural Networks, bag of words form deep-neural-network) network, CNN networks
Or RNN (Recurrent Neural Network, Recognition with Recurrent Neural Network) network etc., the present embodiment is to this and without limit
System.
Picture further feature data, such as these features of the expression generation network video of qualitative character data and click feature data
Physical significance depending on.If orderly form as similar picture, text, can also use CNN or RNN networks, if
It is unordered set feature, uses BoW-DNN networks.
Wherein, the qualitative character of picture includes but is not limited to the website classification of picture, the area stepping of picture.The matter of atlas
Measure feature is using the qualitative character of all pictures of atlas as input, the overall qualitative character of output atlas.Including but not limited to scheme
The average website classification of collection, the average area stepping of atlas.Here, website is classified, and area stepping is directly proportional to quality.
The click feature of atlas browses completeness adjustment on the basis of the click feature of atlas front cover, and according to atlas,
Browse the high atlas of completeness and obtain " reward ", the low atlas of completeness is by " punishment ".
The original deep-neural-network built in advance is trained using training sample including：
Choose the training sample of setting quantity；
The training sample specifically includes：By training search term, and positive sample corresponding with the training search term difference
The positive and negative training pair that this picture and atlas and negative sample picture and atlas are formed；
A training sample is obtained successively to input into the original deep-neural-network, and according to the original deep layer god
The output result based on the training sample through network, the weighting parameters in the original deep-neural-network are adjusted；
Specifically include：
By it is described training search term and data input corresponding with the positive sample picture and atlas to it is described original
In deep-neural-network structure identical first network, and obtain the first predicted value of the first network output；
By the training search term and data input corresponding with the negative sample picture and atlas to described first
In the network of network structure identical second, and obtain the second predicted value of the second network output；
According to first predicted value, second predicted value and the positive sample picture and atlas and the negative sample
Correlation partial order between picture and atlas, counting loss function；
Setting right value update algorithm is taken, along the direction for minimizing loss function, reversely successively updates first net
The weighting parameters of each layer in network and second network.
Judge whether to reach training termination condition set in advance：If so, the original deep layer nerve completed will be trained
Network is as the order models；Otherwise, returning to execution, one training sample of acquisition is inputted to the original deep layer nerve successively
In network, and according to the output result of the original deep-neural-network based on the training sample, to the original deep layer god
It is adjusted through the weighting parameters in network, until reaching training termination condition set in advance.
In the present embodiment, training termination condition can be set according to the actual requirements, for example, training rounds (for example,
1000 times, either 2000 is inferior) or neutral net to aggregated error value of training sample etc., the present embodiment to this and without limit
System.
In 104, the search result items after the sequence are shown.Specifically,
By the thumbnail mixing of picture and atlas, it is illustrated in by hybrid-sorting order on result of page searching.
Preferably, atlas shows most important, most representative one or number in such picture by the way of stacking
Pictures.The content for being both to embody such main picture using this stacking ways of presentation purpose, web page display is saved again
Space, moreover it is possible to give people imitate reality in place pictorial manner aesthetic feeling.
When selecting the atlas on result of page searching, such as when mouse or other dynamic input devices are moved to it
In an atlas region on when, this atlas will be considered as wish by user understand atlas, so should show
More detailed situation.The atlas on the cursor region is showed into the state of activation that is defined, i.e., shared by the atlas
Regional location is significantly greater than other classifications, while the picture overlapped way that the category is included occurs slowly to change, such as with
The mode of animation causes the picture on upper strata to be slowly moved to lower floor, and the picture of lower floor is sequentially moved to top layer by stacking, allows use
Family has an opportunity to watch the picture being blocked in the past due to space limitation.
Preferably, atlas is used using the form that numeral is marked on thumbnail, to represent the picture number in the atlas
Mesh.
The technical scheme provided using above-described embodiment, it can avoid not using search engine technique pair in the prior art
Atlas builds storehouse and retrieval ordering, it is necessary to excavates atlas resource, scalability and ageing all poor for high frequency search in advance；
The feature for not introducing atlas dimension participates in sequence, it is impossible to the effectively correlation and quality of control atlas；It is special not introduce click
Sign forms negative feedback mechanism, and low-quality atlas can not be in system the problem of natural subsidence.User can be reduced in search result
The cost of figure is selected, while meets the needs of user is to complete picture.
It should be noted that for foregoing each method embodiment, in order to be briefly described, therefore it is all expressed as a series of
Combination of actions, but those skilled in the art should know, the application is not limited by described sequence of movement because
According to the application, some steps can use other orders or carry out simultaneously.Secondly, those skilled in the art should also know
Know, embodiment described in this description belongs to preferred embodiment, involved action and module not necessarily the application
It is necessary.
In the described embodiment, the description to each embodiment all emphasizes particularly on different fields, and does not have the portion being described in detail in some embodiment
Point, it may refer to the associated description of other embodiment.
Fig. 5 is the schematic flow sheet for the picture searching device that the embodiment of the application one provides, as shown in figure 5, including following
Module：
Receiving module 51, for receiving search term；
The search term is encapsulated in Fa send Give search engines in searching request by browser, for asking the search engine
Search picture/the pictures related to the search term.
Search module 52, for obtaining search result items from database according to the search term；Specifically, such as Fig. 6 institutes
Show, including following submodule：
Acquisition submodule 61 is indexed, for before the photographic search engine receives the search term, pre-establishing
Picture inverted index and atlas inverted index in scan for, obtain the index that is matched with the search term；
Preferably, scanned for respectively in the picture inverted index and atlas inverted index pre-established, acquisition and institute
State the index of search term matching.
Search result items generate submodule 62, for obtaining corresponding with the index of the search term matching picture and figure
Collection, generate search result items.
Preferably, picture inverted index and atlas inverted index are established respectively.Wherein, it is non-to establish picture inverted index
It is often ripe, it be not discussed in detail.
Further, the search module also includes the picture inverted index and atlas inverted index setting up submodule
63, for before the search term is received, being captured to network picture, establish picture inverted index and atlas falls to arrange rope
Draw, as shown in fig. 7, specifically including：
Text message acquisition submodule 71, the structured text field for webpage where the picture to crawl are divided
Analysis, obtain the text message of the picture；The structured field of webpage includes the Web page subject of the webpage where the picture
Description field, picture periphery the text field and picture binary content field.
Specifically, cutting word processing is carried out to the Web page subject description field of webpage where the picture, by cutting word processing
As a result the input as key phrases extraction.The descriptor extracted herein, also can be more in addition to usual used word
The compound word of accurate more polynary expression picture semantic, i.e., be made up of two or more collocations.
It is that the result that the text message of above-mentioned picture is carried out to cutting word processing is carried out when wherein extracting word as descriptor
Stop words filters, and extracts the word of default part of speech as descriptor, is typically to extract proper noun as descriptor.
It is to be carried out from the text message of above-mentioned picture at cutting word when extracting the collocation of more than two words as descriptor
Extraction meets the collocation of the two or more word of default Collocation pattern as descriptor in the result of reason.
Atlas generates submodule 72, for being excavated to the picture of crawl, generates atlas；Divide according to the mode of excavation is different
For webpage atlas and intelligence polymerization atlas.Resulting each webpage atlas, intelligence polymerization atlas are combined into atlas resource set.
Include the subject description of atlas, the description of the binary content and picture of picture on the content composition of the atlas
Text (i.e. periphery text of the picture on webpage).
In a kind of implementation of the present embodiment, the picture of crawl is polymerize, obtains intelligence polymerization atlas.It is preferred that
, it is polymerize according to its descriptor.Preferably, skill is contrasted for example, by the picture of the various prior arts of mode identification technology
Art is polymerize.
In another implementation of the present embodiment, same webpage or continuous webpage are given birth to by the good picture of subject editing
Into webpage atlas；Specifically；
A secondary picture is randomly selected from database as the first picture, the search second picture similar to the first picture,
First picture is polymerize atlas with second picture composition intelligence.The database can be stored with searched in advance engine from network (for example,
Internet) collect or crawl picture and picture where webpage.It can be searched for from database and the similarity of the first picture
More than predetermined threshold picture as second picture, for example, can be for example, by the various prior arts of mode identification technology
Picture correlation technique searches for second picture.
Webpage where obtaining each second image from image data base.It should be understood that due to second image has can
It can be present in multiple webpages, therefore, for each second image, at least one webpage can be obtained.
For each webpage, the 3rd image that the link of at least level deep of the webpage is pointed to is obtained.For example, webpage
Link on the chained representation of the first order depth webpage, the net that the link on the second level chained representation of the webpage webpage is pointed to
Link on page, by that analogy.
Preferably, the image in the webpage is obtained as the 3rd image.In addition, when at least level deep of the webpage
Link in exist represent page turning link when, obtain represent page turning link pointed by webpage in image as the 3rd figure
Picture.It can determine that the link indicates whether page turning by descriptive text (for example, page up, lower one page) of link etc..
Area (that is, resolution ratio) is selected to be more than the image of predetermined threshold as the 4th image among the 3rd image.So,
Unessential small figure, corner figure etc. can be filtered.
According to the constitutive characteristic of the network address of the 4th image, the 4th image is grouped, to obtain at least one webpage figure
Collection.
Generally, the most contents of the network address of complete image are identicals, are only that the difference of numbering (for example, network address
Content before last level separator "/" is identical, and content afterwards is different, i.e., described picture be located at unified webpage or
Continuous webpage), therefore the 4th similar image of the constitutive characteristic of network address can be divided into one group, as a webpage atlas.
Atlas inverted index setting up submodule 73, for establishing atlas inverted index；
Preferably, the knot that text can will be described to the Web page subject of webpage where the picture of each atlas do cutting word processing
Descriptor of the fruit as atlas；Establish atlas inverted index.
Order module 53, for the search result items to be carried out with hybrid-sorting, the search result items include described search
Picture and atlas and its URL corresponding to the index of rope word matching；The search result items are carried out hybrid-sorting i.e. to picture and
Atlas carries out hybrid-sorting；As shown in figure 8, including following submodule：
Feature extraction submodule 81, for carrying out the feature extraction of atlas dimension, picture and the atlas data can wrap
Include：Atlas text (picture periphery text/atlas subject description) data, image content data, qualitative character data, click feature
Data.
During the hybrid-sorting of atlas and picture, it is preferred that atlas feature is alignd with picture feature, if there is
Corresponding atlas feature, uses the feature of atlas；If there is no corresponding atlas feature, then using the feature of atlas head figures
Alignment；Picture and atlas are ranked up according to the feature after the alignment of picture and atlas.
Hybrid-sorting submodule 82, for the input of the characteristic of term and picture to be sorted and atlas to be instructed in advance
Experienced order models, hybrid-sorting is carried out to respectively picture and atlas to be sorted.
Specifically, the sequence mould for the characteristic of term and picture to be sorted and atlas to be inputted to training in advance
Type, relativity measurement value that can respectively between picture and atlas to be sorted and the term；Based on the correlation degree
Value, respectively picture and atlas to be sorted are subjected to hybrid-sorting.
The order models are deep-neural-network, and the original deep-neural-network built in advance is entered using training sample
Row training obtains.
The original deep-neural-network includes：Represent vector generation network and correlation computations network, it is described represent to
Amount generation network is used to different types of data in the training sample being converted to expression vector and inputted to the correlometer
Network is calculated, the correlation computations network is used to the expression vector of input being converted to a relativity measurement value；The training sample
This includes term and picture and atlas characteristic.
In a preferred embodiment of the present embodiment,
The correlation computations network can include：Hidden layer collection and the output being connected with the output end of the hidden layer collection
Layer；Wherein, the hidden layer collection includes one or more end to end hidden layer, the expression of the vector generation network to
Amount output end is connected with the input of the hidden layer collection, and the output layer exports the relativity measurement value.
Because picture and atlas characteristic include：Atlas text (picture periphery text/atlas subject description) data, figure
Piece content-data, qualitative character data, click feature data.It is corresponding, it is described to represent to include five in vector generation network
Vectorial generation unit is represented, is respectively used to associate in the term of input, atlas text data, image content data and picture
Characteristic is converted to corresponding expression vector, is worked with carrying out follow-up model training.
Wherein, it is described to represent vectorial generation unit, a variety of implementations can be had according to the difference of task object：
The expression vector generation of image content data at present using it is wide be CNN (Convolutional Neural
Network, convolutional neural networks) sorter network, the input of the network is the normalized picture pixels matrix of size, is exported as figure
The class probability distribution of piece represents vector, and classification expression vector is usually picture in a picture category complicated variant system (figure classification system
The general class label for having thousand grades to ten thousand grades) on class probability distribution vector A1,A2,…,An.Wherein, A=(A1,A2,…,
An)；Ai(i=1,2 ..., n) is that the picture that CNN networks provide belongs to the probability of i-th of classification, and n is the size (class of classification system
Other number).
Because atlas text data and term are text, therefore the two represents vectorial generating mode phase one
Cause, be the expression vector generation of text.
Text first passes through participle, then each participle according to default dictionary be mapped as an one-hot (solely heat) characterize to
Amount.Such as：(..., 0 ..., 1 ..., 0 ...), the vector length is the size of dictionary, and it is 1 to have an element, and remaining element is whole
For 0, the position number where element 1 corresponds to sequence number of the word in dictionary.Ensuing processing can have several selections, example
Such as BoW-DNN (Bag of Words-Deep Neural Networks, bag of words form deep-neural-network) network, CNN networks
Or RNN (Recurrent Neural Network, Recognition with Recurrent Neural Network) network etc., the present embodiment is to this and without limit
System.
Picture further feature data, such as these features of the expression generation network video of qualitative character data and click feature data
Physical significance depending on.If orderly form as similar picture, text, can also use CNN or RNN networks, if
It is unordered set feature, uses BoW-DNN networks.
Wherein, the qualitative character of picture includes but is not limited to the website classification of picture, the area stepping of picture.The matter of atlas
Measure feature is using the qualitative character of all pictures of atlas as input, the overall qualitative character of output atlas.Including but not limited to scheme
The average website classification of collection, the average area stepping of atlas.Here, website is classified, and area stepping is directly proportional to quality.
The click feature of atlas browses completeness adjustment on the basis of the click feature of atlas front cover, and according to atlas,
Browse the high atlas of completeness and obtain " reward ", the low atlas of completeness is by " punishment ".
The original deep-neural-network built in advance is trained using training sample including：
Choose the training sample of setting quantity；The training sample specifically includes：By training search term, and with the instruction
Practice the positive and negative training pair that positive sample picture and atlas and negative sample picture and atlas corresponding to search term difference are formed；
A training sample is obtained successively to input into the original deep-neural-network, and according to the original deep layer god
The output result based on the training sample through network, the weighting parameters in the original deep-neural-network are adjusted；
Specifically include：
By it is described training search term and data input corresponding with the positive sample picture and atlas to it is described original
In deep-neural-network structure identical first network, and obtain the first predicted value of the first network output；
By the training search term and data input corresponding with the negative sample picture and atlas to described first
In the network of network structure identical second, and obtain the second predicted value of the second network output；
According to first predicted value, second predicted value and the positive sample picture and atlas and the negative sample
Correlation partial order between picture and atlas, counting loss function；
Setting right value update algorithm is taken, along the direction for minimizing loss function, reversely successively updates first net
The weighting parameters of each layer in network and second network.
Judge whether to reach training termination condition set in advance：If so, the original deep layer nerve completed will be trained
Network is as the order models；Otherwise, returning to execution, one training sample of acquisition is inputted to the original deep layer nerve successively
In network, and according to the output result of the original deep-neural-network based on the training sample, to the original deep layer god
It is adjusted through the weighting parameters in network, until reaching training termination condition set in advance.
In the present embodiment, training termination condition can be set according to the actual requirements, for example, training rounds (for example,
1000 times, either 2000 is inferior) or neutral net to aggregated error value of training sample etc., the present embodiment to this and without limit
System.
Display module 54, for showing the search result items after the sequence；Specifically,
By the thumbnail mixing of atlas and picture, it is illustrated in by hybrid-sorting order on result of page searching.
Preferably, atlas shows most important, most representative one or number in such picture by the way of stacking
Pictures.The content for being both to embody such main picture using this stacking ways of presentation purpose, web page display is saved again
Space, moreover it is possible to give people imitate reality in place pictorial manner aesthetic feeling.
When selecting the atlas on result of page searching, such as when mouse or other dynamic input devices are moved to it
In an atlas region on when, this atlas will be considered as wish by user understand atlas, so should show
More detailed situation.The atlas on the cursor region is showed into the state of activation that is defined, i.e., shared by the atlas
Regional location is significantly greater than other classifications, while the picture overlapped way that the category is included occurs slowly to change, such as with
The mode of animation causes the picture on upper strata to be slowly moved to lower floor, and the picture of lower floor is sequentially moved to top layer by stacking, allows use
Family has an opportunity to watch the picture being blocked in the past due to space limitation.
Preferably, atlas is used using the form that numeral is marked on thumbnail, to represent the picture number in the atlas
Mesh.
The technical scheme provided using above-described embodiment, it can avoid not using search engine technique pair in the prior art
Atlas builds storehouse and retrieval ordering, it is necessary to excavates atlas resource, scalability and ageing all poor for high frequency search in advance；
The feature for not introducing atlas dimension participates in sequence, it is impossible to the effectively correlation and quality of control atlas；It is special not introduce click
Sign forms negative feedback mechanism, and low-quality atlas can not be in system the problem of natural subsidence.User can be reduced in search result
The cost of figure is selected, while meets the needs of user is to complete picture.
It is apparent to those skilled in the art that for convenience and simplicity of description, the terminal of the description
With the specific work process of server, the corresponding process in preceding method embodiment is may be referred to, will not be repeated here.
In several embodiments provided herein, it should be understood that disclosed method and apparatus, it can be passed through
Its mode is realized.For example, device embodiment described above is only schematical, for example, the division of the unit, only
Only a kind of division of logic function, there can be other dividing mode when actually realizing, such as multiple units or component can be tied
Another system is closed or is desirably integrated into, or some features can be ignored, or do not perform.It is another, it is shown or discussed
Mutual coupling or direct-coupling or communication connection can be the INDIRECT COUPLINGs or logical by some interfaces, device or unit
Letter connection, can be electrical, mechanical or other forms.
The unit illustrated as separating component can be or may not be physically separate, show as unit
The part shown can be or may not be physical location, you can with positioned at a place, or can also be distributed to multiple
On NE.Some or all of unit therein can be selected to realize the mesh of this embodiment scheme according to the actual needs
's.
In addition, each functional unit in each embodiment of the application can be integrated in a processing unit, can also
That unit is individually physically present, can also two or more units it is integrated in a unit.The integrated list
Member can both be realized in the form of hardware, can also be realized in the form of hardware adds SFU software functional unit.
Fig. 9 shows the frame suitable for being used for the exemplary computer system/server 012 for realizing embodiment of the present invention
Figure.The computer system/server 012 that Fig. 9 is shown is only an example, function that should not be to the embodiment of the present invention and use
Range band carrys out any restrictions.
As shown in figure 9, computer system/server 012 is showed in the form of universal computing device.Computer system/clothes
The component of business device 012 can include but is not limited to：One or more processor or processing unit 016, system storage
028, the bus 018 of connection different system component (including system storage 028 and processing unit 016).
Bus 018 represents the one or more in a few class bus structures, including memory bus or Memory Controller,
Peripheral bus, graphics acceleration port, processor or the local bus using any bus structures in a variety of bus structures.Lift
For example, these architectures include but is not limited to industry standard architecture (ISA) bus, MCA (MAC)
Bus, enhanced isa bus, VESA's (VESA) local bus and periphery component interconnection (PCI) bus.
Computer system/server 012 typically comprises various computing systems computer-readable recording medium.These media can be appointed
The usable medium what can be accessed by computer system/server 012, including volatibility and non-volatile media, movably
With immovable medium.
System storage 028 can include the computer system readable media of form of volatile memory, such as deposit at random
Access to memory (RAM) 030 and/or cache memory 032.Computer system/server 012 may further include other
Removable/nonremovable, volatile/non-volatile computer system storage medium.Only as an example, storage system 034 can
For reading and writing immovable, non-volatile magnetic media (Fig. 9 is not shown, is commonly referred to as " hard disk drive ").Although in Fig. 9
Being not shown, can providing for the disc driver to may move non-volatile magnetic disk (such as " floppy disk ") read-write, and pair can
The CD drive of mobile anonvolatile optical disk (such as CD-ROM, DVD-ROM or other optical mediums) read-write.In these situations
Under, each driver can be connected by one or more data media interfaces with bus 018.Memory 028 can include
At least one program product, the program product have one group of (for example, at least one) program module, and these program modules are configured
To perform the function of various embodiments of the present invention.
Program/utility 040 with one group of (at least one) program module 042, can be stored in such as memory
In 028, such program module 042 includes --- but being not limited to --- operating system, one or more application program, other
Program module and routine data, the realization of network environment may be included in each or certain combination in these examples.Journey
Sequence module 042 generally performs function and/or method in embodiment described in the invention.
Computer system/server 012 can also with one or more external equipments 014 (such as keyboard, sensing equipment,
Display 024 etc.) communication, in the present invention, computer system/server 012 is communicated with outside radar equipment, can also be with
One or more enables a user to the equipment communication interacted with the computer system/server 012, and/or with causing the meter
Any equipment that calculation machine systems/servers 012 can be communicated with one or more of the other computing device (such as network interface card, modulation
Demodulator etc.) communication.This communication can be carried out by input/output (I/O) interface 022.Also, computer system/clothes
Being engaged in device 012 can also be by network adapter 020 and one or more network (such as LAN (LAN), wide area network (WAN)
And/or public network, such as internet) communication.As illustrated, network adapter 020 by bus 018 and computer system/
Other modules communication of server 012.It should be understood that although not shown in Fig. 3, computer system/server 012 can be combined
Using other hardware and/or software module, include but is not limited to：Microcode, device driver, redundant processing unit, outside magnetic
Dish driving array, RAID system, tape drive and data backup storage system etc..
Processing unit 016 is stored in the program in system storage 028 by operation, described in the invention so as to perform
Function and/or method in embodiment.
Above-mentioned computer program can be arranged in computer-readable storage medium, i.e., the computer-readable storage medium is encoded with
Computer program, the program by one or more computers when being performed so that one or more computers are performed in the present invention
State the method flow shown in embodiment and/or device operation.
Over time, the development of technology, medium implication is more and more extensive, and the route of transmission of computer program is no longer limited by
Tangible medium, directly can also be downloaded from network etc..Any combination of one or more computer-readable media can be used.
Computer-readable medium can be computer-readable signal media or computer-readable recording medium.Computer-readable storage medium
Matter for example may be-but not limited to-system, device or the device of electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or
Combination more than person is any.The more specifically example (non exhaustive list) of computer-readable recording medium includes：With one
Or the electrical connections of multiple wires, portable computer diskette, hard disk, random access memory (RAM), read-only storage (ROM),
Erasable programmable read only memory (EPROM or flash memory), optical fiber, portable compact disc read-only storage (CD-ROM), light
Memory device, magnetic memory device or above-mentioned any appropriate combination.In this document, computer-readable recording medium can
Be it is any include or the tangible medium of storage program, the program can be commanded execution system, device or device use or
Person is in connection.
Computer-readable signal media can include in a base band or as carrier wave a part propagation data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including --- but
It is not limited to --- electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be
Any computer-readable medium beyond computer-readable recording medium, the computer-readable medium can send, propagate or
Transmit for by instruction execution system, device either device use or program in connection.
The program code included on computer-readable medium can be transmitted with any appropriate medium, including --- but it is unlimited
In --- wireless, electric wire, optical cable, RF etc., or above-mentioned any appropriate combination.
It can be write with one or more programming languages or its combination for performing the computer that operates of the present invention
Program code, described program design language include object oriented program language-such as Java, Smalltalk, C++,
Also include conventional procedural programming language-such as " C " language or similar programming language.Program code can be with
Fully perform, partly perform on the user computer on the user computer, the software kit independent as one performs, portion
Divide and partly perform or performed completely on remote computer or server on the remote computer on the user computer.
Be related in the situation of remote computer, remote computer can pass through the network of any kind --- including LAN (LAN) or
Wide area network (WAN) is connected to subscriber computer, or, it may be connected to outer computer (such as provided using Internet service
Business passes through Internet connection).
Finally it should be noted that：Above example is only to illustrate the technical scheme of the application, rather than its limitations；Although
The application is described in detail with reference to the foregoing embodiments, it will be understood by those within the art that：It still may be used
To be modified to the technical scheme described in foregoing embodiments, or equivalent substitution is carried out to which part technical characteristic；
And these modification or replace, do not make appropriate technical solution essence depart from each embodiment technical scheme of the application spirit and
Scope.
Claims (16)
- A kind of 1. image searching method, it is characterised in that including：Receive search term；Search result items are obtained from picture database and atlas database according to the search term；Hybrid-sorting is carried out to the search result items；Show the search result items after the sequence.
- 2. image searching method according to claim 1, it is characterised in that it is described according to the search term from image data Search result items are obtained in storehouse and atlas database includes following sub-step：Scanned in the picture inverted index and atlas inverted index pre-established, obtain the rope matched with the search term Draw；Corresponding with the index of the search term matching picture and atlas are obtained, generates search result items.
- 3. image searching method according to claim 2, it is characterised in that before the search term is received, to network Picture is captured, and establishes inverted index, including following sub-step：The structured text field of webpage where the picture of crawl is analyzed, obtains the text message of the picture；The picture of crawl is excavated, generates atlas；Establish atlas inverted index.
- 4. image searching method according to claim 3, it is characterised in that the picture of described pair of crawl excavates, raw Include into atlas：The picture of crawl is polymerize, obtains intelligence polymerization atlas.
- 5. image searching method according to claim 3, it is characterised in that the picture of described pair of crawl excavates, raw Include into atlas：By the constitutive characteristic of network address is similar and descriptor identical picture generation webpage atlas
- 6. image searching method according to claim 1, it is characterised in that described that the search result items are mixed Sequence further comprises following sub-step：Feature extraction is carried out to picture and atlas；The characteristic of term and picture to be sorted and atlas is inputted to the order models of training in advance, to respectively treating ordering chart Piece and atlas carry out hybrid-sorting.
- 7. image searching method according to claim 6, it is characterised in that the search result after the displaying sequence Item includes：The thumbnail of picture and atlas is illustrated on result of page searching by hybrid-sorting order.
- A kind of 8. picture searching device, it is characterised in that including：Receiving module, for receiving search term；Search module, for obtaining search result items from picture database and atlas database according to the search term；Order module, for carrying out hybrid-sorting to the search result items；Display module, for showing the search result items after the sequence.
- 9. picture searching device according to claim 8, it is characterised in that the search module includes following submodule：Acquisition submodule is indexed, for being scanned in the picture inverted index and atlas inverted index pre-established, is obtained The index matched with the search term；Search result items generate submodule, for obtaining corresponding with the index of the search term matching picture and atlas, generation Search result items.
- 10. picture searching device according to claim 9, it is characterised in that the search module also includes inverted index Setting up submodule, it is described for before the search term is received, being captured to network picture, establishing atlas inverted index Inverted index setting up submodule includes：Text message acquisition submodule, the structured text field for webpage where the picture to crawl are analyzed, and are obtained The text message of the picture；Atlas generates submodule, for being excavated to the picture of crawl, generates atlas；Atlas inverted index setting up submodule, for establishing atlas inverted index.
- 11. picture searching device according to claim 10, it is characterised in that the atlas generation submodule is specifically held OK：The picture of crawl is polymerize, obtains intelligence polymerization atlas.
- 12. picture searching device according to claim 10, it is characterised in that the atlas generation submodule is specifically held OK：By the constitutive characteristic of network address is similar and descriptor identical picture generation webpage atlas
- 13. picture searching device according to claim 8, it is characterised in that the order module includes following submodule：Feature extraction submodule, for carrying out feature extraction to picture and atlas；The hybrid-sorting submodule, for the characteristic of term and picture to be sorted and atlas to be inputted into training in advance Order models, hybrid-sorting is carried out to respectively picture and atlas to be sorted.
- 14. picture searching device according to claim 13, it is characterised in that the display module specifically performs：The thumbnail of picture and atlas is illustrated on result of page searching by hybrid-sorting order.
- 15. a kind of equipment, it is characterised in that the equipment includes：One or more processors；Storage device, for storing one or more programs,When one or more of programs are by one or more of computing devices so that one or more of processors are real The now method as described in any in claim 1-7.
- 16. a kind of computer-readable recording medium, is stored thereon with computer program, it is characterised in that the program is by processor The method as described in any in claim 1-7 is realized during execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710527201.0A CN107451194A (en) | 2017-06-30 | 2017-06-30 | A kind of image searching method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710527201.0A CN107451194A (en) | 2017-06-30 | 2017-06-30 | A kind of image searching method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107451194A true CN107451194A (en) | 2017-12-08 |
Family
ID=60487638
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710527201.0A Pending CN107451194A (en) | 2017-06-30 | 2017-06-30 | A kind of image searching method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107451194A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918184A (en) * | 2019-03-01 | 2019-06-21 | 腾讯科技（深圳）有限公司 | Picture processing system, method and relevant apparatus and equipment |
| CN113190698A (en) * | 2021-04-28 | 2021-07-30 | 北京百度网讯科技有限公司 | Paired picture set generation method and device, electronic equipment and storage medium |
| WO2023155746A1 (en) * | 2022-02-16 | 2023-08-24 | 华为技术有限公司 | Picture search method and related apparatus |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1952935A (en) * | 2006-09-22 | 2007-04-25 | 南京搜拍信息技术有限公司 | Search system and technique comprehensively using information of graphy and character |
| CN101398832A (en) * | 2007-09-30 | 2009-04-01 | 国际商业机器公司 | Image searching method and system by utilizing human face detection |
| CN101901249A (en) * | 2009-05-26 | 2010-12-01 | 复旦大学 | Text-based query expansion and sort method in image retrieval |
| US20110106805A1 (en) * | 2009-10-30 | 2011-05-05 | International Business Machines Corporation | Method and system for searching multilingual documents |
| CN103927328A (en) * | 2014-03-18 | 2014-07-16 | 清华大学 | Query intention mining method and system |
| CN104462590A (en) * | 2014-12-30 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Information searching method and device |
| CN106708940A (en) * | 2016-11-11 | 2017-05-24 | 百度在线网络技术（北京）有限公司 | Method and device used for processing pictures |
-
2017
- 2017-06-30 CN CN201710527201.0A patent/CN107451194A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1952935A (en) * | 2006-09-22 | 2007-04-25 | 南京搜拍信息技术有限公司 | Search system and technique comprehensively using information of graphy and character |
| CN101398832A (en) * | 2007-09-30 | 2009-04-01 | 国际商业机器公司 | Image searching method and system by utilizing human face detection |
| CN101901249A (en) * | 2009-05-26 | 2010-12-01 | 复旦大学 | Text-based query expansion and sort method in image retrieval |
| US20110106805A1 (en) * | 2009-10-30 | 2011-05-05 | International Business Machines Corporation | Method and system for searching multilingual documents |
| CN103927328A (en) * | 2014-03-18 | 2014-07-16 | 清华大学 | Query intention mining method and system |
| CN104462590A (en) * | 2014-12-30 | 2015-03-25 | 百度在线网络技术（北京）有限公司 | Information searching method and device |
| CN106708940A (en) * | 2016-11-11 | 2017-05-24 | 百度在线网络技术（北京）有限公司 | Method and device used for processing pictures |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918184A (en) * | 2019-03-01 | 2019-06-21 | 腾讯科技（深圳）有限公司 | Picture processing system, method and relevant apparatus and equipment |
| CN109918184B (en) * | 2019-03-01 | 2023-09-26 | 腾讯科技（深圳）有限公司 | Picture processing system, method and related device and equipment |
| CN113190698A (en) * | 2021-04-28 | 2021-07-30 | 北京百度网讯科技有限公司 | Paired picture set generation method and device, electronic equipment and storage medium |
| CN113190698B (en) * | 2021-04-28 | 2023-08-01 | 北京百度网讯科技有限公司 | Paired picture set generation method and device, electronic equipment and storage medium |
| WO2023155746A1 (en) * | 2022-02-16 | 2023-08-24 | 华为技术有限公司 | Picture search method and related apparatus |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111062871B (en) | Image processing method and device, computer equipment and readable storage medium | |
| US9449271B2 (en) | Classifying resources using a deep network | |
| CN112215171B (en) | Target detection method, device, equipment and computer readable storage medium | |
| CN111310041B (en) | Image-text publishing method, model training method and device and storage medium | |
| CN114419509B (en) | Multi-mode emotion analysis method and device and electronic equipment | |
| CN113434716B (en) | Cross-modal information retrieval method and device | |
| CN115982376B (en) | Method and device for training model based on text, multimode data and knowledge | |
| CN109165316A (en) | A kind of method for processing video frequency, video index method, device and terminal device | |
| CN115131698B (en) | Video attribute determining method, device, equipment and storage medium | |
| CN112487242A (en) | Method and device for identifying video, electronic equipment and readable storage medium | |
| CN116601626A (en) | Personal knowledge graph construction method and device and related equipment | |
| CN115455171B (en) | Text video mutual inspection rope and model training method, device, equipment and medium | |
| CN114372414A (en) | Multi-modal model construction method and device and computer equipment | |
| CN107451194A (en) | A kind of image searching method and device | |
| CN109933217A (en) | Method and apparatus for pushing sentence | |
| CN107944026A (en) | A kind of method, apparatus, server and the storage medium of atlas personalized recommendation | |
| CN110264277A (en) | Data processing method and device, medium and the calculating equipment executed by calculating equipment | |
| CN110222144A (en) | Method for extracting content of text, device, electronic equipment and storage medium | |
| CN115438225B (en) | Video text mutual inspection method and model training method, device, equipment and medium thereof | |
| CN106446696A (en) | Information processing method and electronic device | |
| KR102358195B1 (en) | System for providing selected articles using linear regression | |
| CN115168609A (en) | Text matching method and device, computer equipment and storage medium | |
| EP4280085A1 (en) | Devices, systems, and methods for displaying and linking legal content | |
| CN116955591A (en) | Recommendation language generation method, related device and medium for content recommendation | |
| CN110222189A (en) | Method and apparatus for output information |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171208 |