CN107609195A - Question searching method and device - Google Patents
Question searching method and device Download PDFInfo
- Publication number
- CN107609195A CN107609195A CN201710971686.2A CN201710971686A CN107609195A CN 107609195 A CN107609195 A CN 107609195A CN 201710971686 A CN201710971686 A CN 201710971686A CN 107609195 A CN107609195 A CN 107609195A
- Authority
- CN
- China
- Prior art keywords
- target photo
- picture
- clear
- text
- graphic pictures
- 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
- 238000000034 method Methods 0.000 title claims abstract description 30
- 238000003062 neural network model Methods 0.000 claims abstract description 41
- 238000005194 fractionation Methods 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 6
- 238000012015 optical character recognition Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The embodiment of the invention discloses a method and a device for searching questions. The method comprises the following steps: acquiring a target picture; if the target picture is a fuzzy picture, splitting the target picture into a text question stem picture and a graphic picture; acquiring a clear text picture corresponding to the text question stem picture based on a preset neural network model and the text question stem picture; acquiring a clear graphic picture corresponding to the graphic picture based on a gradient algorithm and the graphic picture; according to the technical scheme, the problem that when a user shoots a picture when using a question searching APP, due to personal operation habits, the shot image is fuzzy, the recognition accuracy is very low, and the searching accuracy is low is solved, and the success rate of question searching can be improved.
Description
Technical field
The present embodiments relate to searching topic field, more particularly to one kind searches topic method and device.
Background technology
At present, intelligent terminal class product has many searching for solution students' work problems to inscribe APP, searches topic APP and passes through to user
Insoluble problem is taken pictures, then topic APP is searched in the content input that will take pictures, and is arrived to search topic APP according to content search of taking pictures
Corresponding answer of solving a problem.
But due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that shooting
Image obscure, so using conventional OCR (Optical Character Recognition, optical character identification) sides
On the basis of method, recognition accuracy is very low, in turn results in the problem of search accuracy rate is low.
The content of the invention
The embodiment of the present invention provides one kind and searches topic method and device, and topic success rate is searched to improve.
In a first aspect, the embodiments of the invention provide one kind to search topic method, including:
Obtain Target Photo;
If the Target Photo is blurred picture, the Target Photo is split as text stem picture and graphic diagram
Piece;
Obtained based on default neural network model and the text stem picture clear corresponding to the text stem picture
Clear textual image;
Clear graphic pictures corresponding to the graphic pictures are obtained based on gradient algorithm and the graphic pictures;
Matching parsing data are obtained according to the clear textual image and the clear graphic pictures.
Further, the acquisition Target Photo includes:
The original image of camera shooting is obtained, wherein, the original image comprises at least object content;
The Target Photo for including object content in the original image is intercepted according to the object content.
Further, if the Target Photo is blurred picture, the Target Photo is split as text stem picture
Include with graphic pictures:
If the absolute value of the difference of the pixel value of the two neighboring pixel in the Target Photo is more than first threshold, and
Less than Second Threshold, it is determined that the Target Photo is blurred picture, wherein, the first threshold is less than Second Threshold;
The Target Photo is split as text stem picture and graphic pictures.
Further, the default neural network model is AlexNet models.
Further, before obtaining Target Photo, in addition to:
Neural network model is established, training, shape are optimized according to neural network model described in default training sample set pair
Into the default neural network model.
Second aspect, the embodiment of the present invention additionally provide one kind and search topic device, and the device includes:
Target Photo acquisition module, for obtaining Target Photo;
Module is split, if being blurred picture for the Target Photo, the Target Photo is split as text stem
Picture and graphic pictures;
Textual image acquisition module, described in being obtained based on default neural network model and the text stem picture
Clear textual image corresponding to text stem picture;
Graphic pictures acquisition module, it is corresponding for obtaining the graphic pictures based on gradient algorithm and the graphic pictures
Clear graphic pictures;
Data acquisition module is parsed, for being obtained therewith according to the clear textual image and the clear graphic pictures
The parsing data matched somebody with somebody.
Further, the Target Photo acquisition module includes:
Original image acquiring unit, for obtaining the original image of camera shooting, wherein, the original image at least wraps
Include object content;
Interception unit, for intercepting the target figure for including object content in the original image according to the object content
Piece.
Further, module is split to be specifically used for:
If the absolute value of the difference of the pixel value of the two neighboring pixel in the Target Photo is more than first threshold, and
Less than Second Threshold, it is determined that the Target Photo is blurred picture, wherein, the first threshold is less than Second Threshold;
The Target Photo is split as text stem picture and graphic pictures.
Further, the default neural network model is AlexNet models.
Further, in addition to:
Model training module, for before Target Photo is obtained, establishing neural network model, according to default training sample
Neural network model described in set pair optimizes training, forms the default neural network model.
The embodiment of the present invention is by obtaining Target Photo;If Target Photo is blurred picture, Target Photo is split as
Text stem picture and graphic pictures;Text stem picture pair is obtained based on default neural network model and text stem picture
The clear textual image answered;Clear graphic pictures corresponding to graphic pictures are obtained based on gradient algorithm and graphic pictures;According to
Clear textual image and clear graphic pictures obtain matching parsing data, by technical scheme, solve
Due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that the image of shooting obscures,
Cause recognition accuracy very low, in turn result in the problem of search accuracy rate is low, it is possible to increase search topic success rate.
Brief description of the drawings
Fig. 1 is a kind of flow chart for searching topic method in the embodiment of the present invention one;
Fig. 2 is a kind of flow chart for searching topic method in the embodiment of the present invention two;
Fig. 3 is a kind of flow chart for searching topic method in the embodiment of the present invention three;
Fig. 4 is a kind of structural representation for searching topic device in the embodiment of the present invention four.
Embodiment
The present invention is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining the present invention, rather than limitation of the invention.It also should be noted that in order to just
Part related to the present invention rather than entire infrastructure are illustrate only in description, accompanying drawing.
Embodiment one
Fig. 1 is a kind of flow chart for searching topic method that the embodiment of the present invention one provides, and the present embodiment is applicable to terminal and set
The standby situation for searching topic, this method can search topic device to perform by provided in an embodiment of the present invention, the device can use software and/
Or the mode of hardware is realized, the device can be integrated in any need and search in the terminal device of topic, such as typically intelligent terminal
(such as smart mobile phone or tablet personal computer etc.), as shown in figure 1, this method specifically comprises the following steps:
S110, obtain Target Photo.
Wherein, the Target Photo is to be shot by the camera installed on the terminal device, when user opens intelligence
In terminal searching topic application after, receive search topic instruction when, open camera photographic subjects picture.Camera can be intelligent hand
The rear camera of machine, the embodiment of the present invention are not limited to the concrete form of camera.
Wherein, the Target Photo comprises at least user needs search answer by what the camera on terminal device was shot
Topic, such as can be examination question A and half of examination question B on user's exercise, wherein examination question A is the topic for needing to search for answer.
Wherein, the mode for obtaining Target Photo can be that the center alignment of camera is needed to the topic of the target topic shot
It is dry to carry out shooting acquisition, or directly to pass through the Target Photo of stem picture of the camera shooting comprising target topic.
S120, if Target Photo is blurred picture, Target Photo is split as text stem picture and graphic pictures.
Wherein, the blurred picture is unsharp picture, i.e., the adjacent pixel values difference comparsion of Target Photo is balanced.
Wherein, judge whether Target Photo is that the mode of blurred picture can be the pixel of pixel in acquisition Target Photo
Value, it is blurred picture that Target Photo is illustrated if the pixel value of the neighbor pixel of Target Photo is more balanced.
Wherein, the Target Photo includes text stem picture and graphic pictures, according to the feature of text and figure by institute
State Target Photo and be split as text stem picture and graphic pictures.
Wherein, the text stem picture is the picture of the textual portions in stem;Graphic pictures are the figure in stem
Partial picture.
Wherein, the graphic pictures are the picture for the necessary figure that can accurately search topic answer.
Specifically, if Target Photo is blurred picture, Target Photo is split as to picture and the figure portion of textual portions
The picture divided.
S130, clear text corresponding to text stem picture is obtained based on default neural network model and text stem picture
This picture.
Wherein, the default neural network model pre-establishes, and inputs to obscure text stem picture, exports to be clear
The model of clear textual image, such as can be, using AlexNet models.
Optionally, the default neural network model is convolutional neural networks model.
Optionally, before obtaining Target Photo, in addition to:
Neural network model is established, training, shape are optimized according to neural network model described in default training sample set pair
Into the default neural network model.
Optionally, the default neural network model is AlexNet models.
Specifically, the structural model of the AlexNet models is described by taking a cpu server as an example.The model has altogether
It is divided into eight layers, 5 Ge Juan basic units, and 3 full articulamentums, excitation function RELU and part are contained in each convolutional layer
Response normalization (LRN) processing, then passing through down-sampled (pool processing).
Specifically, the neural network model that fuzzy text stem picture input is trained, exports fuzzy text topic
Clear textual image corresponding to dry picture.
S140, clear graphic pictures corresponding to graphic pictures are obtained based on gradient algorithm and graphic pictures.
Wherein, the gradient algorithm is a kind of Local Optimization Algorithm, for for graphic pictures to be converted into clear graphic diagram
The method of piece, such as gradient descent method.
Specifically, graphic pictures are converted to by clear graphic pictures based on gradient algorithm.
S150, matching parsing data are obtained according to clear textual image and clear graphic pictures.
Wherein, the parsing data can be the answer of topic corresponding to Target Photo, or Target Photo is corresponding
Topic thinking of solving a problem, such as can be, if entitled multiple-choice question corresponding to Target Photo, topic corresponding with Target Photo
The parsing data of mesh matching can be the option of selection, or how obtain the course of solving questions of option.
Specifically, the mode that matching parsing data are obtained according to clear textual image and clear graphic pictures can be with
Looked into for graphical information corresponding to the text message according to corresponding to clear textual image acquisition textual image and clear graphic pictures
Look for matching parsing data.
The technical scheme of the present embodiment, by obtaining Target Photo;If Target Photo is blurred picture, by Target Photo
It is split as text stem picture and graphic pictures;Text stem is obtained based on default neural network model and text stem picture
Clear textual image corresponding to picture;Clear graphic diagram corresponding to graphic pictures is obtained based on gradient algorithm and graphic pictures
Piece;Matching parsing data are obtained according to clear textual image and clear graphic pictures, by technical scheme,
Solve due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that the figure of shooting
As obscuring, cause recognition accuracy very low, in turn result in the problem of search accuracy rate is low, it is possible to increase search topic success rate.
Embodiment two
Fig. 2 is a kind of flow chart for searching topic method in the embodiment of the present invention two, and the present embodiment is with previous embodiment one
Basis optimizes, there is provided preferably searches topic method, is specifically, the acquisition Target Photo includes:Obtain camera shooting
Original image, wherein, the original image comprises at least object content;The original image is intercepted according to the object content
In include the Target Photo of object content.
Accordingly, the method for the present embodiment specifically comprises the following steps:
S210, the original image of camera shooting is obtained, wherein, original image comprises at least object content.
Wherein, the camera is the camera on intelligent terminal, and topic is searched when user is opened in intelligent terminal
After, receive search topic instruction when, open camera.Such as can be the rear camera of smart mobile phone, the present invention is implemented
Example is not limited to the concrete form of camera.
Wherein, the mode of camera shooting original image can be that the center alignment of camera is needed into the topic shot
Stem, or the picture directly obtained by stem of the camera shooting comprising topic.
Wherein, the object content is the stem content for needing to search for the topic of answer, such as can be, if object content
For topic A, then original image may include topic A and other contents unrelated with topic A.
Specifically, when user open search topic application after, open intelligent terminal on camera, by camera shoot to
Include the original image of object content less.
S220, the Target Photo for including object content in original image is intercepted according to object content.
Specifically, according to object content intercept original image in comprising object content Target Photo mode can be
When user shoots, the center alignment in the visual field is directly needed into the topic shot, then can intercept original image intermediate region
The stem picture of complete topic is Target Photo.
S230, if Target Photo is blurred picture, Target Photo is split as text stem picture and graphic pictures.
S240, clear text corresponding to text stem picture is obtained based on default neural network model and text stem picture
This picture.
S250, clear graphic pictures corresponding to graphic pictures are obtained based on gradient algorithm and graphic pictures.
S260, matching parsing data are obtained according to clear textual image and clear graphic pictures.
In a specific example, the original image of object content is included by camera shooting, from the original graph
Interception includes the Target Photo of object content in piece, and the Target Photo is converted into gray scale picture, extracts all offices of picture
Portion's feature, the characteristic value in match index, the picture most matched is found, and then obtain the answer of topic corresponding to Target Photo.
The technical scheme of the present embodiment, by obtaining Target Photo;If Target Photo is blurred picture, by Target Photo
It is split as text stem picture and graphic pictures;Text stem is obtained based on default neural network model and text stem picture
Clear textual image corresponding to picture;Clear graphic diagram corresponding to graphic pictures is obtained based on gradient algorithm and graphic pictures
Piece;Matching parsing data are obtained according to clear textual image and clear graphic pictures, by technical scheme,
Solve due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that the figure of shooting
As obscuring, cause recognition accuracy very low, in turn result in the problem of search accuracy rate is low, it is possible to increase search topic success rate.
Embodiment three
Fig. 3 is a kind of flow chart for searching topic method in the embodiment of the present invention three, and the present embodiment is using previous embodiment as base
Plinth optimizes, there is provided preferably searches topic method, is specifically, if the Target Photo is blurred picture, by the target
Picture, which is split as text stem picture and graphic pictures, to be included:If the pixel value of the two neighboring pixel in the Target Photo
The absolute value of difference be more than first threshold, and be less than Second Threshold, it is determined that the Target Photo is blurred picture, wherein,
The first threshold is less than Second Threshold;The Target Photo is split as text stem picture and graphic pictures.
Accordingly, the method for the present embodiment specifically comprises the following steps:
S310, obtain Target Photo.
S320, if the absolute value of the difference of the pixel value of two neighboring pixel in Target Photo is more than first threshold,
And it is less than Second Threshold, it is determined that Target Photo is blurred picture, wherein, first threshold is less than Second Threshold.
Wherein, the first threshold and Second Threshold can be that user sets, or the data rule of thumb obtained.
The first threshold and Second Threshold are primarily to judge the difference of the pixel value of the two neighboring pixel in Target Photo
It is whether balanced.Such as can be that first threshold can be 10, Second Threshold can be 20.
S330, Target Photo is split as text stem picture and graphic pictures.
S340, clear text corresponding to text stem picture is obtained based on default neural network model and text stem picture
This picture.
S350, clear graphic pictures corresponding to graphic pictures are obtained based on gradient algorithm and graphic pictures.
S360, matching parsing data are obtained according to clear textual image and clear graphic pictures.
In a specific example, when user shoots picture, due to personal operating habit, it may appear that shooting
Image is obscured, and carries out searching topic by fuzzy picture, and search accuracy rate is low, and network establishes process, and fuzzy promotion picture is cut
For text promotion and figure, the training pattern (using AlexNet models) of text is established respectively, pattern algorithm uses gradient method
It is acquired, is trained by more than one hundred million fuzzy promotion pictures.After user, which opens, searches topic application, by intelligent terminal
Camera take pictures, cut it is fuzzy promote picture, by the blurred picture upload server of cutting, blurred picture, which is pushed to, has instructed
The network perfected.
The technical scheme of the present embodiment, by obtaining Target Photo;If Target Photo is blurred picture, by Target Photo
It is split as text stem picture and graphic pictures;Text stem is obtained based on default neural network model and text stem picture
Clear textual image corresponding to picture;Clear graphic diagram corresponding to graphic pictures is obtained based on gradient algorithm and graphic pictures
Piece;Matching parsing data are obtained according to clear textual image and clear graphic pictures, by technical scheme,
Solve due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that the figure of shooting
As obscuring, cause recognition accuracy very low, in turn result in the problem of search accuracy rate is low, it is possible to increase search topic success rate.
Example IV
Fig. 4 is a kind of structural representation for searching topic device of the embodiment of the present invention four.The present embodiment is applicable to terminal and set
The standby situation for searching topic, the device can realize that the device can be integrated in any offer and search topic work(by the way of software and/or hardware
In the equipment of energy, such as typically intelligent terminal (such as smart mobile phone or tablet personal computer etc.), as shown in figure 4, described search topic
Device specifically includes:Target Photo acquisition module 410, split module 420, textual image acquisition module 430, graphic pictures acquisition
Module 440 and parsing data acquisition module 450.
Wherein, Target Photo acquisition module 410, for obtaining Target Photo;
Module 420 is split, if being blurred picture for the Target Photo, the Target Photo is split as text topic
Dry picture and graphic pictures;
Textual image acquisition module 430, for based on default neural network model and text stem picture acquisition
Clear textual image corresponding to the text stem picture;
Graphic pictures acquisition module 440, for obtaining the graphic pictures based on gradient algorithm and the graphic pictures
Corresponding clear graphic pictures;
Parse data acquisition module 450, for according to the clear textual image and the clear graphic pictures obtain with
Matching parsing data.
Optionally, the Target Photo acquisition module includes:
Original image acquiring unit, for obtaining the original image of camera shooting, wherein, the original image at least wraps
Include object content;
Interception unit, for intercepting the target figure for including object content in the original image according to the object content
Piece.
Optionally, module is split to be specifically used for:
If the absolute value of the difference of the pixel value of the two neighboring pixel in the Target Photo is more than first threshold, and
Less than Second Threshold, it is determined that the Target Photo is blurred picture, wherein, the first threshold is less than Second Threshold;
The Target Photo is split as text stem picture and graphic pictures.
Optionally, the default neural network model is AlexNet models.
Optionally, in addition to:
Model training module, for before Target Photo is obtained, establishing neural network model, according to default training sample
Neural network model described in set pair optimizes training, forms the default neural network model.
The technical scheme of the present embodiment, by obtaining Target Photo;If Target Photo is blurred picture, by Target Photo
It is split as text stem picture and graphic pictures;Text stem is obtained based on default neural network model and text stem picture
Clear textual image corresponding to picture;Clear graphic diagram corresponding to graphic pictures is obtained based on gradient algorithm and graphic pictures
Piece;Matching parsing data are obtained according to clear textual image and clear graphic pictures, by technical scheme,
Solve due to use search topic APP when, user shoot picture when, due to personal operating habit, it may appear that the figure of shooting
As obscuring, cause recognition accuracy very low, in turn result in the problem of search accuracy rate is low, it is possible to increase search topic success rate.
The said goods can perform the method that any embodiment of the present invention is provided, and possess the corresponding functional module of execution method
And beneficial effect.
Pay attention to, above are only presently preferred embodiments of the present invention and institute's application technology principle.It will be appreciated by those skilled in the art that
The invention is not restricted to specific embodiment described here, can carry out for a person skilled in the art various obvious changes,
Readjust and substitute without departing from protection scope of the present invention.Therefore, although being carried out by above example to the present invention
It is described in further detail, but the present invention is not limited only to above example, without departing from the inventive concept, also
Other more equivalent embodiments can be included, and the scope of the present invention is determined by scope of the appended claims.
Claims (10)
1. one kind searches topic method, it is characterised in that including:
Obtain Target Photo;
If the Target Photo is blurred picture, the Target Photo is split as text stem picture and graphic pictures;
Clear text corresponding to the text stem picture is obtained based on default neural network model and the text stem picture
This picture;
Clear graphic pictures corresponding to the graphic pictures are obtained based on gradient algorithm and the graphic pictures;
Matching parsing data are obtained according to the clear textual image and the clear graphic pictures.
2. according to the method for claim 1, it is characterised in that the acquisition Target Photo includes:
The original image of camera shooting is obtained, wherein, the original image comprises at least object content;
The Target Photo for including object content in the original image is intercepted according to the object content.
3. according to the method for claim 1, it is characterised in that if the Target Photo is blurred picture, by the mesh
Piece of marking on a map, which is split as text stem picture and graphic pictures, to be included:
If the absolute value of the difference of the pixel value of the two neighboring pixel in the Target Photo is more than first threshold, and is less than
Second Threshold, it is determined that the Target Photo is blurred picture, wherein, the first threshold is less than Second Threshold;
The Target Photo is split as text stem picture and graphic pictures.
4. according to the method for claim 1, it is characterised in that the default neural network model is AlexNet models.
5. according to the method for claim 1, it is characterised in that before obtaining Target Photo, in addition to:
Neural network model is established, training is optimized according to neural network model described in default training sample set pair, forms institute
State default neural network model.
6. one kind searches topic device, it is characterised in that including:
Target Photo acquisition module, for obtaining Target Photo;
Module is split, if being blurred picture for the Target Photo, the Target Photo is split as text stem picture
And graphic pictures;
Textual image acquisition module, for obtaining the text based on default neural network model and the text stem picture
Clear textual image corresponding to stem picture;
Graphic pictures acquisition module is clear corresponding to the graphic pictures for being obtained based on gradient algorithm and the graphic pictures
Clear graphic pictures;
Data acquisition module is parsed, it is matching for being obtained according to the clear textual image and the clear graphic pictures
Parse data.
7. device according to claim 6, it is characterised in that the Target Photo acquisition module includes:
Original image acquiring unit, for obtaining the original image of camera shooting, wherein, the original image comprises at least mesh
Mark content;
Interception unit, for intercepting the Target Photo for including object content in the original image according to the object content.
8. device according to claim 6, it is characterised in that the fractionation module is specifically used for:
If the absolute value of the difference of the pixel value of the two neighboring pixel in the Target Photo is more than first threshold, and is less than
Second Threshold, it is determined that the Target Photo is blurred picture, wherein, the first threshold is less than Second Threshold;
The Target Photo is split as text stem picture and graphic pictures.
9. device according to claim 6, it is characterised in that the default neural network model is AlexNet models.
10. device according to claim 6, it is characterised in that also include:
Model training module, for before Target Photo is obtained, establishing neural network model, according to default training sample set pair
The neural network model optimizes training, forms the default neural network model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710971686.2A CN107609195A (en) | 2017-10-18 | 2017-10-18 | Question searching method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710971686.2A CN107609195A (en) | 2017-10-18 | 2017-10-18 | Question searching method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107609195A true CN107609195A (en) | 2018-01-19 |
Family
ID=61078521
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710971686.2A Pending CN107609195A (en) | 2017-10-18 | 2017-10-18 | Question searching method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107609195A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109492644A (en) * | 2018-10-16 | 2019-03-19 | 深圳壹账通智能科技有限公司 | A kind of matching and recognition method and terminal device of exercise image |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103700121A (en) * | 2013-12-30 | 2014-04-02 | Tcl集团股份有限公司 | Method and device for compressing composite image |
CN104463103A (en) * | 2014-11-10 | 2015-03-25 | 小米科技有限责任公司 | Image processing method and device |
CN104715253A (en) * | 2015-04-02 | 2015-06-17 | 北京贞观雨科技有限公司 | Method and server for obtaining test question analysis information |
CN106548169A (en) * | 2016-11-02 | 2017-03-29 | 重庆中科云丛科技有限公司 | Fuzzy literal Enhancement Method and device based on deep neural network |
-
2017
- 2017-10-18 CN CN201710971686.2A patent/CN107609195A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103700121A (en) * | 2013-12-30 | 2014-04-02 | Tcl集团股份有限公司 | Method and device for compressing composite image |
CN104463103A (en) * | 2014-11-10 | 2015-03-25 | 小米科技有限责任公司 | Image processing method and device |
CN104715253A (en) * | 2015-04-02 | 2015-06-17 | 北京贞观雨科技有限公司 | Method and server for obtaining test question analysis information |
CN106548169A (en) * | 2016-11-02 | 2017-03-29 | 重庆中科云丛科技有限公司 | Fuzzy literal Enhancement Method and device based on deep neural network |
Non-Patent Citations (2)
Title |
---|
杨杰等: ""视频目标检测和跟踪及其应用"", 《视频目标检测和跟踪及其应用》 * |
洪汉玉: ""现代图像图形处理与分析"", 《现代图像图形处理与分析》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109492644A (en) * | 2018-10-16 | 2019-03-19 | 深圳壹账通智能科技有限公司 | A kind of matching and recognition method and terminal device of exercise image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10936897B2 (en) | Method and system for information extraction from document images using conversational interface and database querying | |
CN111190939A (en) | User portrait construction method and device | |
CN103824053A (en) | Face image gender marking method and face gender detection method | |
US20230119593A1 (en) | Method and apparatus for training facial feature extraction model, method and apparatus for extracting facial features, device, and storage medium | |
CN110363084A (en) | A kind of class state detection method, device, storage medium and electronics | |
CN113704531A (en) | Image processing method, image processing device, electronic equipment and computer readable storage medium | |
CN107679070B (en) | Intelligent reading recommendation method and device and electronic equipment | |
Surikov et al. | Floor plan recognition and vectorization using combination unet, faster-rcnn, statistical component analysis and ramer-douglas-peucker | |
CN103678460B (en) | For identifying the method and system for the non-text elements for being suitable to be communicated in multi-language environment | |
CN113761105A (en) | Text data processing method, device, equipment and medium | |
CN115331150A (en) | Image recognition method, image recognition device, electronic equipment and storage medium | |
CN111753120A (en) | Method and device for searching questions, electronic equipment and storage medium | |
CN104615639B (en) | A kind of method and apparatus for providing the presentation information of picture | |
Pertusa et al. | MirBot: A collaborative object recognition system for smartphones using convolutional neural networks | |
CN108810577B (en) | User portrait construction method and device and electronic equipment | |
Hung et al. | Automatic vietnamese passport recognition on android phones | |
CN112749769B (en) | Graphic code detection method, graphic code detection device, computer equipment and storage medium | |
CN107609195A (en) | Question searching method and device | |
CN116958729A (en) | Training of object classification model, object classification method, device and storage medium | |
CN114332882A (en) | Text translation method and device, electronic equipment and storage medium | |
WO2016155537A1 (en) | Method and device for ranking search results of picture objects | |
CN113377980A (en) | Information labeling method and device, electronic equipment and storage medium | |
Rahul et al. | Deep reader: Information extraction from document images via relation extraction and natural language | |
CN111881338A (en) | Printed matter content retrieval method based on social software light application applet | |
CN112307244A (en) | Photographic picture screening system based on image significance detection and human eye state detection |
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 |
Application publication date: 20180119 |
|
RJ01 | Rejection of invention patent application after publication |