CN108829855A - It is worn based on the clothing that condition generates confrontation network and takes recommended method, system and medium - Google Patents
It is worn based on the clothing that condition generates confrontation network and takes recommended method, system and medium Download PDFInfo
- Publication number
- CN108829855A CN108829855A CN201810646047.3A CN201810646047A CN108829855A CN 108829855 A CN108829855 A CN 108829855A CN 201810646047 A CN201810646047 A CN 201810646047A CN 108829855 A CN108829855 A CN 108829855A
- Authority
- CN
- China
- Prior art keywords
- network
- clothing
- layer
- clothing image
- generates
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
It is worn the invention discloses the clothing for generating confrontation network based on condition and takes recommended method, system and medium, including:True clothing image data set is established, each picture in data set is labeled with corresponding attribute tags;Confrontation net structure, which is generated, based on condition generates network G;Attribute tags are input to and are generated in network G, output clothing image pattern;Confrontation net structure, which is generated, based on condition differentiates network D;It by true clothing the image data true clothing image concentrated and obtained clothing image pattern while being input in differentiation network D, the attribute tags of output clothing image and the true and false judging result for wearing image clothes;Alternating iteration training differentiates network D and generates network G;Attribute tags content to be recommended is received, generates to wear clothes to wear using trained generation network G and takes picture.
Description
Technical field
The invention belongs to depth learning technology field, wears to take more particularly to the clothing for generating confrontation network based on condition and push away
Recommend method, system and medium.
Background technique
Deep learning is the key areas of machine learning, it can complete to need the artificial intelligence of high abstraction feature to appoint
Business, it made breakthrough progress in the application of the multiclass such as voice, image recognition and retrieval, natural language understanding in recent years.
Generating confrontation network (GAN) is the rising star in nearly 2 years deep learning fields, it is by generation network and differentiates that network is constituted, its benefit
With " two-person game " thought, updates two networks respectively by back-propagation algorithm to execute competitive study and reach trained mesh
's.Thus according to the thought for generating confrontation network, it derives many variants and is applied well.Generate confrontation network
(GAN) characterization learnt can be used for a variety of applications, including image synthesis, semantic image editor, Style Transfer, image super-resolution
Technology and classification etc..
With the development of economy and society, suitable clothing wear take it is more important.The many existing clothing in market, which are worn, at present takes
Software is mainly recommended according to gender, season and style, and clothing is all with purchase link mostly, and software is mostly with sale
For the purpose of profit.However, when not premised on buying clothes, people want according to the weather on the same day (as rained, fine day) and fit
Existing clothes of being arranged in pairs or groups with occasion (such as dancing party is gone to school) becomes problem.Even dependence parent also in need is come dress of arranging in pairs or groups
Young student possibly according to factors such as weather can not suitably wear and take when leaving parent.
Summary of the invention
In order to solve the deficiencies in the prior art, is worn the present invention provides the clothing for generating confrontation network based on condition and take recommendation
Method, system and medium can be effectively combined user demand and obtain suitably wearing clothes to wear taking recommendation picture.
As the first aspect of the present invention, the clothing for generating confrontation network based on condition is provided to wear and take recommended method;
To achieve the goals above, the technical solution adopted by the present invention is as follows:
It is worn based on the clothing that condition generates confrontation network and takes recommended method, including:
Step (1):True clothing image data set is established, each picture in data set is labeled with corresponding attribute
Label;
Step (2):Confrontation net structure, which is generated, based on condition generates network G;The attribute tags of step (1) are input to life
At in network G, image pattern is worn in output clothes;
Step (3):Confrontation net structure, which is generated, based on condition differentiates network D;By the true clothing image data of step (1)
The obtained clothing image pattern of true clothing image and step (2) concentrated is input to simultaneously to be differentiated in network D, and output clothing is schemed
The attribute tags of picture and the true and false judging result of clothing image;
Step (4):Alternating iteration training differentiates network D and generates network G;
Step (5):Attribute tags content to be recommended is received, generates to wear clothes to wear using trained generation network G and takes figure
Piece.
As a further improvement of the present invention, the step of step (1) is:
Step (101):Acquire the unified whole clothing picture comprising jacket and lower clothing of background color;According to setting pixel
All clothing pictures are normalized in size;
Step (102):Corresponding attribute tags are marked for each clothing picture;The attribute tags include:Gender, day
Gas and occasion.
As a further improvement of the present invention, the step of step (2) are:
The network structure for generating network G uses super-resolution depth residual error network model SRResNet, generates the G's of network
Network structure specifically includes 39 layers:
First layer is full articulamentum, and input generates the characteristic patterns of 1024 channel 16*16 resolution ratio after first layer, the 2nd,
5,8,11,14,17,20,23,28,32,36 layers are crowd normalization layer BN, the 4th, 7,10,13,16,19,22,26,30,34,38
Layer is convolutional layer Conv, and the 3rd, 6,12,18,24,29,33,37 layer is active coating Relu, and the 9th, 15,21,25 layer is to ask by element
With layer Elementwise Sum, the 27th, 31,35 layer is shuffled a layer Pixel Shuffler for pixel, and the 39th layer is output layer.
The attribute tags of each picture in random noise signal and step (1) in data set are input to generation network
In G, output clothing image pattern;
As a further improvement of the present invention, the step of step (3) are:
Differentiate that the network structure of the D of network specifically includes 72 layers:
1st, 3,5,8,10,13,15,17,21,23,27,29,31,35,37,41,43,45,49,51,55,57,59,
63,65,69 layers are convolutional layer Conv,
2nd, 4,7,9,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,
50,52,54,56,58,60,62,64,66,68,70 layers are Leaky active coating Leaky Relu,
6th, 11,19,25,33,39,47,53,61,67 layer is the layer Elementwise Sum that sums by element,
71st, 72 layer is full articulamentum;
By the true clothing image that the true clothing image data of step (1) is concentrated and the clothing image that step (2) obtains
Sample is input to simultaneously to be differentiated in network D, the attribute tags of output clothing image and the true and false judging result of clothing image;
As a further improvement of the present invention, the step of step (4) are:
Step (401):Network D total loss function is differentiated when setting training;
Step (402):Network G total loss function is generated when setting training;
Step (403):Update the parameter for differentiating network D;
Step (404):The parameter of more newly-generated network G;
Step (405):Independent alternating iteration training differentiates network D and generates network G, repeat step (403) and
(404), until reaching the number of iterations of setting.
The step of step (401) is:Differentiate the total loss function of network D by the first Classification Loss function, first pair
Anti- loss function and gradient punishment loss function composition, differentiate the total loss function L (D) of network D:
L (D)=Lcls(D)+λadvLadv(D)+λgpLgp(D) (1)
Wherein, LclsIt (D) is the first Classification Loss function for differentiating network D, Ladv(D) the first confrontation to differentiate network D
Loss function, LgpIt (D) is the gradient punishment loss function for differentiating network D, λadvTo differentiate that network D's fights the flat of loss function
Weigh the factor, and λ is arrangedadvEqual to the quantity of attribute tags, λgpFor the balance factor of the gradient penalty of differentiation network D;
Wherein, differentiate the first Classification Loss function L of network Dcls(D):
Wherein, PdataIndicate the distribution for really wearing image data set x obtained in step (1) clothes, PnoiseExpression is made an uproar at random
The distribution of acoustical signal z, PcondExpression has distributed the prior distribution of label c,For the first Classification Loss
[logP in functionD[labelx| x]] expectation,For in the first Classification Loss function
log(PD[c | G (z, c)]) expectation;
Wherein, differentiate the first confrontation loss function L of network Dadv(D):
Wherein,The expectation for fighting logD (x) in loss function for first,For the expectation of log (1-D (G (z, c))) in the first confrontation loss function;
Wherein, differentiate the gradient penalty L of network Dgp(D):
Wherein, PpertubeddataIndicate the distribution of interference data,Letter is punished for gradient
In numberExpectation.
The step of step (402) is:
Network G total loss function is generated when training to be made of the second Classification Loss function and the second confrontation loss function,
It is as follows to generate the total loss function of network G:
L (G)=Lcls(G)+λadv′Ladv(G) (5)
Wherein, Lcls(G) the second Classification Loss function of network G, L are made a living intoadv(G) second pair of damage-retardation of network is made a living into
Lose function, λadv' make a living into network confrontation loss function balance factor, be arranged λadv' equal to the quantity of attribute tags;
Wherein, the second Classification Loss function L of network G is generatedcls(G) as follows:
Wherein, PnoiseIndicate the distribution of random noise signal z, PcondExpression has distributed the prior distribution of label c,For log (P in the second Classification Loss functionD[c | G (z, c)]) expectation;
Wherein, the second confrontation loss function L of network G is generatedadv(G) as follows:
Wherein,Indicate the expectation of log (D (G (z, c))) in the second confrontation loss function;
The step of step (403) is:
True clothing image data set x that step (1) obtains and corresponding attribute tags are input to and differentiated in network D,
According to the logP in the first Classification Loss function for differentiating network DD[labelx| x] part, first confrontation loss function in
LogD (x) punishes loss function with gradient partially to update the parameter for differentiating network D;The fixed parameter for differentiating network D, will be random
Noise signal z and 3 dimension attribute label vector c are input to generation network G, then wear the generation for generating network G output clothes image
Sample, which is input to, to be differentiated in network D, further according to the log (P in the first Classification Loss function for differentiating network DD[c|G(z,c)])
Partially, the log (1-D (G (z, c))) in the first confrontation loss function partially punishes loss function with gradient to update differentiation network
The parameter of D.
The step of step (404) is:The model parameter of the fixed differentiation network D obtained by step (403), will be random
Noise signal z and 3 dimension attribute label vector c, which is input to, to be generated in network G, according to the Classification Loss function and life for generating network G
Carry out the parameter of more newly-generated network G at the confrontation loss function of network G.
As a further improvement of the present invention, the step of step (5) are:
The customized 3 dimension attribute label of random noise signal and user is input to trained generation network G, generates net
Network can generate the clothing image of corresponding attribute.
Such as inputting 3 dimension attribute labels is " female, fine day, dancing party ", can generate corresponding clothing image, thus user can incite somebody to action
The clothing of customized generation, which is worn, takes image as reference, come the clothing that oneself same day of arranging in pairs or groups is applicable.
As a second aspect of the invention, the clothing for generating confrontation network based on condition is provided to wear and take recommender system;
To achieve the goals above, the technical solution adopted by the present invention is as follows:
It is worn based on the clothing that condition generates confrontation network and takes recommender system, including:It memory, processor and is stored in
The computer instruction run on reservoir and on a processor when the computer instruction is run by processor, is completed any of the above-described
Step described in method.
As the third aspect of the present invention, a kind of computer readable storage medium is provided;
To achieve the goals above, the technical solution adopted by the present invention is as follows:
A kind of computer readable storage medium, is stored thereon with computer instruction, and the computer instruction is transported by processor
When row, step described in any of the above-described method is completed.
Compared with prior art, the beneficial effects of the invention are as follows:
The present invention generates confrontation network model using based on DRAGAN network model come structural environment, and the model is than other
Calculation amount needed for part generates confrontation network model is relatively fewer, and restrains comparatively fast during model training, can make to generate
Network quickly generates more stable generation image;
The demand that the present invention can effectively provide user, which is integrated into, to be come, and is with sale different from the market most of
Purpose and be that leading clothing is worn and takes software with clothes style, user only needs the three attribute " property of customed clothing image
Not ", " weather " and " occasion ", the clothing that can obtain generation network output quickly, which is worn, takes image, and as with reference to next
The existing clothes of user is arranged in pairs or groups out proper applicable dress.
Detailed description of the invention
The accompanying drawings constituting a part of this application is used to provide further understanding of the present application, and the application's shows
Meaning property embodiment and its explanation are not constituted an undue limitation on the present application for explaining the application.
Fig. 1 is holistic approach flow chart of the invention;
Fig. 2 (a)-Fig. 2 (c) makes a living into the network structure of network G;
Fig. 3 (a)-Fig. 3 (c) is the network structure for differentiating network D.
Specific embodiment
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific embodiment, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singular
Also it is intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or their combination.
As shown in Figure 1, a kind of clothing for generating confrontation network based on condition is worn and takes recommended method, include the following steps:
(1) true clothing image data set is established;
(2) network G is generated based on DRAGAN network model construction;
(3) network D is differentiated based on DRAGAN network model construction;
(4) training differentiates network D and generates network G;
(5) it is worn using the customized generation clothing of trained generation network G and takes picture.
In the step (1), true clothing image data set is established, is specifically included:
(1-1):The whole clothing picture for collecting a large amount of high quality and the upper and lower clothing of background color unification, to all clothing
Picture carries out size normalization processing, the image that processing pixel is 128 × 128;
(1-2):The attribute for setting up clothing image is three kinds:Gender, weather and occasion, for every clothing picture mark this three
Attribute label, then every picture has corresponding three attribute.
In the step (2), network G is generated based on DRAGAN network model construction, is specifically included:
The input for generating network G is random noise signal z and 3 dimension attribute label vector c, generates the clothing of network G output
Image pattern will be as the input for differentiating network D.
The network structure for generating network G uses super-resolution depth residual error network model SRResNet, including 16 residual values
Block (ResBlocks) and 3 sub-pixel-level convolutional layer CNN for characteristics of image up-sampling, specific structure such as Fig. 2 (a)-figure
Shown in 2 (c), the input for generating network G generates the clothing image of 3 channel 128*128 resolution ratio after generating network G, and will
It is as the input for differentiating network D.
In the step (3), network D is differentiated based on DRAGAN network model construction, is specifically included:
Differentiate that the true clothing of foundation in the clothing image pattern and step (1) of network G generation is made a living into the input of network D
Image pattern, output is for judging to wear the true and false of image clothes and judging the attribute tags of clothing image.Differentiate the network knot of network D
Structure includes 10 residual value blocks (ResBlocks), and all batches of normalization layers (BN) are removed in differentiating network D, and last
Convolutional layer on additionally plus one layer of full articulamentum is as multi-tag classifier, specific structure such as Fig. 3 (a)-Fig. 3 (c) is shown, by step
Suddenly the clothing image that the 3 channel 128*128 resolution ratio that network G generates are generated in (2) exports clothing respectively after differentiating network D
The attribute tags of image and the true and false judging result of clothing image.
In the step (4), training differentiates network D and generates network G, and specific steps include:
(4-1):Differentiate that the total loss function of network D is lost by the first Classification Loss function, the first confrontation when setting training
Function and gradient punishment loss function composition, differentiate that the total loss function of network D is as follows:
L (D)=Lcls(D)+λadvLadv(D)+λgpLgp(D) (1)
Wherein, LclsIt (D) is the Classification Loss function for differentiating network, LadvIt (D) is the confrontation loss function for differentiating network, Lgp
It (D) is the gradient punishment loss function for differentiating network, λadvFor the balance factor for fighting loss function for differentiating network, λ is setadv
Quantity (λ other equal to tag classadv=3), λgpFor the balance factor of the gradient penalty of differentiation network, λ is setgp=0.5;
Wherein, differentiate the Classification Loss function L of network Dcls(D) as follows:
Wherein, PdataIndicate the distribution for really wearing image data set x obtained in step (1) clothes, PnoiseExpression is made an uproar at random
The distribution of acoustical signal z, PcondExpression has distributed the prior distribution of label c,For the first Classification Loss
[logP in functionD[labelx| x]] expectation,For in the first Classification Loss function
log(PD[c | G (z, c)]) expectation;
Wherein, differentiate the confrontation loss function L of network Dadv(D) as follows:
Wherein,The expectation for fighting logD (x) in loss function for first,For the expectation of log (1-D (G (z, c))) in the first confrontation loss function;
Wherein, differentiate the gradient penalty L of network Dgp(D) as follows:
Wherein, PpertubeddataIndicate the distribution of interference data,Letter is punished for gradient
In numberExpectation;
(4-2):Network G total loss function is generated when setting training to be lost by the second Classification Loss function and the second confrontation
It is as follows to generate the total loss function of network G for function composition:
L (G)=Lcls(G)+λadvLadv(G) (5)
Wherein, Lcls(G) the Classification Loss function of network, L are made a living intoadv(G) the confrontation loss function of network is made a living into,
λadvThe balance factor for making a living into the confrontation loss function of network, is arranged λadvQuantity (λ other equal to tag classadv=3);
Wherein, the Classification Loss function L of network G is generatedcls(G) as follows:
Wherein, PnoiseIndicate the distribution of random noise signal z, PcondExpression has distributed the prior distribution of label c,For log (P in the second Classification Loss functionD[c | G (z, c)]) expectation;
Wherein, the confrontation loss function L of network G is generatedadv(G) as follows:
Wherein,Indicate the expectation of log (D (G (z, c))) in the second confrontation loss function;
(4-3):Update the parameter for differentiating network D, the true clothing image data set x that step (1) is obtained and corresponding
Class label, which is input to, to be differentiated in network D, according to the logP in the first Classification Loss function for differentiating network DD[labelx| x] portion
Point, first confrontation loss function in logD (x) partially with gradient punish loss function come update differentiate network D parameter, Gu
Surely the parameter for differentiating network D, is input to generation network G for random noise signal z and 3 dimension attribute label vector c, then will generate
The generation clothing image pattern of network G output, which is input to, to be differentiated in network D, further according to the first Classification Loss letter for differentiating network D
Log (P in numberD[c | G (z, c)]) part, the log (1-D (G (z, c))) in the first confrontation loss function partially punish with gradient
Loss function is penalized to update the parameter for differentiating network D;
(4-4):The parameter of more newly-generated network G, the model parameter of the fixed differentiation network D obtained by step (4-3), will
Random noise signal z and 3 dimension attribute label vector c, which is input to, to be generated in network G, according to the second Classification Loss for generating network G
Function and the second confrontation loss function carry out the parameter of more newly-generated network G;
(4-5):Independent alternating iteration training differentiates network D and generates network G, repeats step (4-3) and (4-4),
Until reaching the number of iterations of setting.
In the step (5), is worn using the customized generation clothing of trained generation network G and take picture, specifically included:It will
Random noise signal and the customized 3 dimension attribute label of user are input to trained generation network G, and generating network can generate pair
The clothing image of attribute is answered, such as inputting 3 dimension attribute labels is " female, fine day, dancing party ", corresponding clothing image can be generated, thus
User, which can wear the clothing of customized generation, takes image as reference, come the clothing that oneself same day of arranging in pairs or groups is applicable.
The foregoing is merely preferred embodiment of the present application, are not intended to limit this application, for the skill of this field
For art personnel, various changes and changes are possible in this application.Within the spirit and principles of this application, made any to repair
Change, equivalent replacement, improvement etc., should be included within the scope of protection of this application.
Claims (8)
1. being worn based on the clothing that condition generates confrontation network and taking recommended method, characterized in that including:
Step (1):True clothing image data set is established, each picture in data set is labeled with corresponding attribute mark
Label;
Step (2):Confrontation net structure, which is generated, based on condition generates network G;The attribute tags of step (1) are input to generation net
In network G, output clothing image pattern;
Step (3):Confrontation net structure, which is generated, based on condition differentiates network D;The true clothing image data of step (1) is concentrated
True clothing image and the obtained clothing image pattern of step (2) while being input to differentiate in network D, output clothing image
The true and false judging result of attribute tags and clothing image;
Step (4):Alternating iteration training differentiates network D and generates network G;
Step (5):Attribute tags content to be recommended is received, generates to wear clothes to wear using trained generation network G and takes picture.
2. the clothing for generating confrontation network based on condition as described in claim 1, which is worn, takes recommended method, characterized in that
The step of step (1) is:
Step (101):Acquire the unified whole clothing picture comprising jacket and lower clothing of background color;According to setting pixel size
All clothing pictures are normalized;
Step (102):Corresponding attribute tags are marked for each clothing picture;The attribute tags include:Gender, weather and
Occasion.
3. the clothing for generating confrontation network based on condition as described in claim 1, which is worn, takes recommended method, characterized in that
The step of step (2) is:
The network structure for generating network G uses super-resolution depth residual error network model SRResNet, generates the network of the G of network
Structure specifically includes 39 layers:
First layer is full articulamentum, and input generates the characteristic patterns of 1024 channel 16*16 resolution ratio after first layer, the 2nd, 5,8,
11,14,17,20,23,28,32,36 layers are crowd normalization layer BN, and the 4th, 7,10,13,16,19,22,26,30,34,38 layer is
Convolutional layer Conv, the 3rd, 6,12,18,24,29,33,37 layer is active coating Relu, and the 9th, 15,21,25 layer is by element summation layer
Elementwise Sum, the 27th, 31,35 layer is shuffled a layer Pixel Shuffler for pixel, and the 39th layer is output layer;
The attribute tags of each picture in random noise signal and step (1) in data set are input to and are generated in network G,
Output clothing image pattern.
4. the clothing for generating confrontation network based on condition as described in claim 1, which is worn, takes recommended method, characterized in that
The step of step (3) is:
Differentiate that the network structure of the D of network specifically includes 72 layers:
1st, 3,5,8,10,13,15,17,21,23,27,29,31,35,37,41,43,45,49,51,55,57,59,63,65,
69 layers are convolutional layer Conv,
2nd, 4,7,9,12,14,16,18,20,22,24,26,28,30,32,34,36,38,40,42,44,46,48,50,52,
54,56,58,60,62,64,66,68,70 layers are Leaky active coating Leaky Relu,
6th, 11,19,25,33,39,47,53,61,67 layer is the layer Elementwise Sum that sums by element,
71st, 72 layer is full articulamentum;
By the true clothing image that the true clothing image data of step (1) is concentrated and the clothing image pattern that step (2) obtains
It is input to and is differentiated in network D simultaneously, the attribute tags of output clothing image and the true and false judging result of clothing image.
5. the clothing for generating confrontation network based on condition as described in claim 1, which is worn, takes recommended method, characterized in that
The step of step (4) is:
Step (401):Network D total loss function is differentiated when setting training;
Step (402):Network G total loss function is generated when setting training;
Step (403):Update the parameter for differentiating network D;
Step (404):The parameter of more newly-generated network G;
Step (405):Independent alternating iteration training differentiates network D and generates network G, repeats step (403) and (404),
Until reaching the number of iterations of setting.
6. the clothing for generating confrontation network based on condition as described in claim 1, which is worn, takes recommended method, characterized in that
The step of step (5) is:
The customized 3 dimension attribute label of random noise signal and user is input to trained generation network G, generates network meeting
Generate the clothing image of corresponding attribute.
7. being worn based on the clothing that condition generates confrontation network and taking recommender system, characterized in that including:Memory, processor and
The computer instruction run on a memory and on a processor is stored, when the computer instruction is run by processor, is completed
Step described in the claims 1-6 either method.
8. a kind of computer readable storage medium, characterized in that be stored thereon with computer instruction, the computer instruction is located
When managing device operation, step described in the claims 1-6 either method is completed.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810646047.3A CN108829855B (en) | 2018-06-21 | 2018-06-21 | Clothing wearing recommendation method, system and medium for generating confrontation network based on condition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810646047.3A CN108829855B (en) | 2018-06-21 | 2018-06-21 | Clothing wearing recommendation method, system and medium for generating confrontation network based on condition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108829855A true CN108829855A (en) | 2018-11-16 |
CN108829855B CN108829855B (en) | 2021-02-19 |
Family
ID=64143166
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810646047.3A Active CN108829855B (en) | 2018-06-21 | 2018-06-21 | Clothing wearing recommendation method, system and medium for generating confrontation network based on condition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108829855B (en) |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109829537A (en) * | 2019-01-30 | 2019-05-31 | 华侨大学 | Style transfer method and equipment based on deep learning GAN network children's garment clothes |
CN110134395A (en) * | 2019-05-17 | 2019-08-16 | 广东工业大学 | A kind of generation method of icon generates system and relevant apparatus |
CN110135032A (en) * | 2019-04-30 | 2019-08-16 | 厦门大学 | A kind of auxiliary clothes generation method and device generating network based on confrontation |
CN110441061A (en) * | 2019-08-13 | 2019-11-12 | 哈尔滨理工大学 | Planet wheel bearing life-span prediction method based on C-DRGAN and AD |
CN110647986A (en) * | 2019-08-13 | 2020-01-03 | 杭州电子科技大学 | Road damage image generation method based on countermeasure type generation network |
CN110909770A (en) * | 2019-11-05 | 2020-03-24 | 上海眼控科技股份有限公司 | ACGAN-combined image sample processing method, apparatus, system, and medium |
CN111476200A (en) * | 2020-04-27 | 2020-07-31 | 华东师范大学 | Face de-identification generation method based on generation of confrontation network |
CN111783980A (en) * | 2020-06-28 | 2020-10-16 | 大连理工大学 | Ranking learning method based on dual cooperation generation type countermeasure network |
CN111794741A (en) * | 2020-08-11 | 2020-10-20 | 中国石油天然气集团有限公司 | Method for realizing sliding directional drilling simulator |
CN112000769A (en) * | 2020-08-17 | 2020-11-27 | 东北林业大学 | Clothing commodity advertisement pattern generation method based on confrontation network |
CN112100908A (en) * | 2020-08-31 | 2020-12-18 | 西安工程大学 | Garment design method for generating confrontation network based on multi-condition deep convolution |
CN112230210A (en) * | 2020-09-09 | 2021-01-15 | 南昌航空大学 | HRRP radar target identification method based on improved LSGAN and CNN |
CN112329116A (en) * | 2020-11-23 | 2021-02-05 | 恩亿科(北京)数据科技有限公司 | Distortion zero space planning design generation method and system based on generation of countermeasure network |
CN112435083A (en) * | 2019-08-26 | 2021-03-02 | 珠海格力电器股份有限公司 | Article recommendation method based on brain wave recognition and electronic equipment |
CN112613445A (en) * | 2020-12-29 | 2021-04-06 | 深圳威富优房客科技有限公司 | Face image generation method and device, computer equipment and storage medium |
CN112927801A (en) * | 2021-01-24 | 2021-06-08 | 武汉东湖大数据交易中心股份有限公司 | Health prediction model construction method and device based on dressing meteorological index |
CN113033595A (en) * | 2020-12-24 | 2021-06-25 | 重庆大学 | Multi-label automobile model generation method based on generation countermeasure network |
CN113555106A (en) * | 2020-04-23 | 2021-10-26 | 浙江远图互联科技股份有限公司 | Intelligent traditional Chinese medicine remote auxiliary diagnosis and treatment platform based on generation countermeasure network |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106803082A (en) * | 2017-01-23 | 2017-06-06 | 重庆邮电大学 | A kind of online handwriting recognition methods based on conditional generation confrontation network |
CN107220600A (en) * | 2017-05-17 | 2017-09-29 | 清华大学深圳研究生院 | A kind of Picture Generation Method and generation confrontation network based on deep learning |
CN107563509A (en) * | 2017-07-17 | 2018-01-09 | 华南理工大学 | A kind of dynamic adjustment algorithm for the condition DCGAN models that feature based returns |
US20180082348A1 (en) * | 2016-09-16 | 2018-03-22 | Conduent Business Services, Llc | Method and system for data processing to recommend list of physical stores in real-time |
US20180101770A1 (en) * | 2016-10-12 | 2018-04-12 | Ricoh Company, Ltd. | Method and system of generative model learning, and program product |
CN107993131A (en) * | 2017-12-27 | 2018-05-04 | 广东欧珀移动通信有限公司 | Wear to take and recommend method, apparatus, server and storage medium |
CN108022213A (en) * | 2017-11-29 | 2018-05-11 | 天津大学 | Video super-resolution algorithm for reconstructing based on generation confrontation network |
CN108109049A (en) * | 2017-12-29 | 2018-06-01 | 广东欧珀移动通信有限公司 | Clothing matching Forecasting Methodology, device, computer equipment and storage medium |
CN108132983A (en) * | 2017-12-14 | 2018-06-08 | 北京小米移动软件有限公司 | The recommendation method and device of clothing matching, readable storage medium storing program for executing, electronic equipment |
-
2018
- 2018-06-21 CN CN201810646047.3A patent/CN108829855B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20180082348A1 (en) * | 2016-09-16 | 2018-03-22 | Conduent Business Services, Llc | Method and system for data processing to recommend list of physical stores in real-time |
US20180101770A1 (en) * | 2016-10-12 | 2018-04-12 | Ricoh Company, Ltd. | Method and system of generative model learning, and program product |
CN106803082A (en) * | 2017-01-23 | 2017-06-06 | 重庆邮电大学 | A kind of online handwriting recognition methods based on conditional generation confrontation network |
CN107220600A (en) * | 2017-05-17 | 2017-09-29 | 清华大学深圳研究生院 | A kind of Picture Generation Method and generation confrontation network based on deep learning |
CN107563509A (en) * | 2017-07-17 | 2018-01-09 | 华南理工大学 | A kind of dynamic adjustment algorithm for the condition DCGAN models that feature based returns |
CN108022213A (en) * | 2017-11-29 | 2018-05-11 | 天津大学 | Video super-resolution algorithm for reconstructing based on generation confrontation network |
CN108132983A (en) * | 2017-12-14 | 2018-06-08 | 北京小米移动软件有限公司 | The recommendation method and device of clothing matching, readable storage medium storing program for executing, electronic equipment |
CN107993131A (en) * | 2017-12-27 | 2018-05-04 | 广东欧珀移动通信有限公司 | Wear to take and recommend method, apparatus, server and storage medium |
CN108109049A (en) * | 2017-12-29 | 2018-06-01 | 广东欧珀移动通信有限公司 | Clothing matching Forecasting Methodology, device, computer equipment and storage medium |
Non-Patent Citations (2)
Title |
---|
CHAO WANG等: "CGAN-PLANKTON: TOWARDS LARGE-SCALE IMBALANCED CLASS GENERATION AND FINE-GRAINED CLASSIFICATION", 《IEEE》 * |
徐一峰等: "生成对抗网络理论模型和应用综述", 《金华职业技术学院学报》 * |
Cited By (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109829537B (en) * | 2019-01-30 | 2023-10-24 | 华侨大学 | Deep learning GAN network children's garment based style transfer method and equipment |
CN109829537A (en) * | 2019-01-30 | 2019-05-31 | 华侨大学 | Style transfer method and equipment based on deep learning GAN network children's garment clothes |
CN110135032A (en) * | 2019-04-30 | 2019-08-16 | 厦门大学 | A kind of auxiliary clothes generation method and device generating network based on confrontation |
CN110134395A (en) * | 2019-05-17 | 2019-08-16 | 广东工业大学 | A kind of generation method of icon generates system and relevant apparatus |
CN110441061A (en) * | 2019-08-13 | 2019-11-12 | 哈尔滨理工大学 | Planet wheel bearing life-span prediction method based on C-DRGAN and AD |
CN110647986A (en) * | 2019-08-13 | 2020-01-03 | 杭州电子科技大学 | Road damage image generation method based on countermeasure type generation network |
CN112435083A (en) * | 2019-08-26 | 2021-03-02 | 珠海格力电器股份有限公司 | Article recommendation method based on brain wave recognition and electronic equipment |
CN110909770A (en) * | 2019-11-05 | 2020-03-24 | 上海眼控科技股份有限公司 | ACGAN-combined image sample processing method, apparatus, system, and medium |
CN113555106A (en) * | 2020-04-23 | 2021-10-26 | 浙江远图互联科技股份有限公司 | Intelligent traditional Chinese medicine remote auxiliary diagnosis and treatment platform based on generation countermeasure network |
CN111476200B (en) * | 2020-04-27 | 2022-04-19 | 华东师范大学 | Face de-identification generation method based on generation of confrontation network |
CN111476200A (en) * | 2020-04-27 | 2020-07-31 | 华东师范大学 | Face de-identification generation method based on generation of confrontation network |
CN111783980A (en) * | 2020-06-28 | 2020-10-16 | 大连理工大学 | Ranking learning method based on dual cooperation generation type countermeasure network |
CN111794741A (en) * | 2020-08-11 | 2020-10-20 | 中国石油天然气集团有限公司 | Method for realizing sliding directional drilling simulator |
CN111794741B (en) * | 2020-08-11 | 2023-08-18 | 中国石油天然气集团有限公司 | Method for realizing sliding directional drilling simulator |
CN112000769A (en) * | 2020-08-17 | 2020-11-27 | 东北林业大学 | Clothing commodity advertisement pattern generation method based on confrontation network |
CN112100908A (en) * | 2020-08-31 | 2020-12-18 | 西安工程大学 | Garment design method for generating confrontation network based on multi-condition deep convolution |
CN112100908B (en) * | 2020-08-31 | 2024-03-22 | 西安工程大学 | Clothing design method for generating countermeasure network based on multi-condition deep convolution |
CN112230210A (en) * | 2020-09-09 | 2021-01-15 | 南昌航空大学 | HRRP radar target identification method based on improved LSGAN and CNN |
CN112329116A (en) * | 2020-11-23 | 2021-02-05 | 恩亿科(北京)数据科技有限公司 | Distortion zero space planning design generation method and system based on generation of countermeasure network |
CN113033595A (en) * | 2020-12-24 | 2021-06-25 | 重庆大学 | Multi-label automobile model generation method based on generation countermeasure network |
CN112613445A (en) * | 2020-12-29 | 2021-04-06 | 深圳威富优房客科技有限公司 | Face image generation method and device, computer equipment and storage medium |
CN112613445B (en) * | 2020-12-29 | 2024-04-30 | 深圳威富优房客科技有限公司 | Face image generation method, device, computer equipment and storage medium |
CN112927801A (en) * | 2021-01-24 | 2021-06-08 | 武汉东湖大数据交易中心股份有限公司 | Health prediction model construction method and device based on dressing meteorological index |
CN112927801B (en) * | 2021-01-24 | 2024-05-14 | 武汉东湖大数据科技股份有限公司 | Construction method and device of health prediction model based on dressing meteorological index |
Also Published As
Publication number | Publication date |
---|---|
CN108829855B (en) | 2021-02-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108829855A (en) | It is worn based on the clothing that condition generates confrontation network and takes recommended method, system and medium | |
Xu et al. | Imagereward: Learning and evaluating human preferences for text-to-image generation | |
You et al. | Relative CNN-RNN: Learning relative atmospheric visibility from images | |
Srinivasan et al. | Biases in generative art: A causal look from the lens of art history | |
Liu et al. | Cones: Concept neurons in diffusion models for customized generation | |
CN109657156A (en) | A kind of personalized recommendation method generating confrontation network based on circulation | |
CN109754317A (en) | Merge interpretation clothes recommended method, system, equipment and the medium of comment | |
CN112308115B (en) | Multi-label image deep learning classification method and equipment | |
Yu et al. | DressUp!: outfit synthesis through automatic optimization. | |
CN112380453B (en) | Article recommendation method and device, storage medium and equipment | |
Li et al. | Retrieving real world clothing images via multi-weight deep convolutional neural networks | |
CN109902912A (en) | A kind of personalized image aesthetic evaluation method based on character trait | |
Zhou et al. | Improved cross-label suppression dictionary learning for face recognition | |
CN110175469A (en) | A kind of social media privacy of user leakage detection method, system, equipment and medium | |
Wang et al. | Proximity-based group formation game model for community detection in social network | |
He et al. | Analysis of the communication method of national traditional sports culture based on deep learning | |
Wang et al. | Learning outfit compatibility with graph attention network and visual-semantic embedding | |
CN110334185A (en) | The treating method and apparatus of data in a kind of platform | |
El Zant et al. | Interactions and influence of world painters from the reduced Google matrix of Wikipedia networks | |
Islam et al. | Learning character design from experts and laymen | |
Mougiakakou et al. | SCAPEVIEWER: preliminary results of a landscape perception classification system based on neural network technology | |
Wang et al. | [Retracted] Design of Sports Training Simulation System for Children Based on Improved Deep Neural Network | |
Tripathi et al. | Facial expression recognition using data mining algorithm | |
CN113947798A (en) | Background replacing method, device and equipment of application program and storage medium | |
CN114283300A (en) | Label determining method and device, and model training method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |