CA3148915A1 - Method and device for automatically generating banner images of a target object in batches - Google Patents
Method and device for automatically generating banner images of a target object in batches Download PDFInfo
- Publication number
- CA3148915A1 CA3148915A1 CA3148915A CA3148915A CA3148915A1 CA 3148915 A1 CA3148915 A1 CA 3148915A1 CA 3148915 A CA3148915 A CA 3148915A CA 3148915 A CA3148915 A CA 3148915A CA 3148915 A1 CA3148915 A1 CA 3148915A1
- Authority
- CA
- Canada
- Prior art keywords
- targct
- target
- thc
- objcct
- master
- 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 46
- 238000012545 processing Methods 0.000 claims abstract description 15
- 239000000463 material Substances 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 19
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000003709 image segmentation Methods 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 4
- 238000004458 analytical method Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000005457 optimization Methods 0.000 claims description 3
- 238000013461 design Methods 0.000 abstract description 8
- 238000004519 manufacturing process Methods 0.000 abstract 1
- 230000003252 repetitive effect Effects 0.000 abstract 1
- 238000004904 shortening Methods 0.000 abstract 1
- 238000004422 calculation algorithm Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 8
- 238000005034 decoration Methods 0.000 description 6
- 238000011156 evaluation Methods 0.000 description 6
- 230000008901 benefit Effects 0.000 description 4
- 239000003086 colorant Substances 0.000 description 3
- 238000004590 computer program Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013210 evaluation model Methods 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Image Analysis (AREA)
- Information Transfer Between Computers (AREA)
Abstract
A method and device for automatically generating banner images of a target object in batches. The method comprises: processing a batch of target object images, and acquiring a target object main body corresponding to each target object image and target object image information (S1); matching, according to the target object main body and the target object image information and from a master set library, at least two target master sets corresponding to each target object main body (S2); acquiring layer information of each target master set, acquiring, according to the layer information, an element corresponding to each layer of each target master set, and determining, according to a layer of each grayed target master set, a matching color of written copy corresponding to the layer (S3); and generating a batch of banner images according to the target object main body, the target object image information, the at least two target master sets, the element corresponding to each layer, the written copy, and the matching color of the written copy (S4). Banner images are automatically generated in batches according to the target object images, thereby shortening a design period, avoiding repetitive operation, and improving the efficiency of manufacturing a banner image.
Description
METHOD AND DEVICE FOR AUTOMATICALLY GENERATING BANNER IMAGES OF A
TARGET OBJECT IN BATCHES
Technical field The present invention is related to the field of image processing technologies, in particular, to a method and device for automatically generating banner images of a target object in batches.
Background With the rapid development and prevalence of intcrnct technologies, online shopping becomes an important part for more and more people. Each c-commerce platforms decorate the product images and add promotion texts to promote information and attract customers, improving commodity transaction rates.
Therefore, generating a high-quality banner image plays a significant role.
Before image processing technologies were applied to the field of visual design, the banner generation is maj orly completed by designer. In the process, designs spend a great amount of time on simple tasks, such as text modifications and spot expansions, wherein the massive redundant work brings limited benefits to designers' skill development. In the meanwhile, precise marketing is the further demand for e-commerce. With the background of massive volume, the poster spots on the home page are required to display in adherence with the principle of "thousands of people-recommendation" (wherein "thousands of people-recommendation" is a recommendation algorithm that can construct customer preference models based on the massive database of e-commerce websites to promote products according to features matching with customer preferences; and the contents displayed on the website read by target customers, to identify potential buyers for sellers for precision marketing). Hereby, high poster generation efficiency is also required.
Therefore, a new banner image generation method is demanded to address the aforementioned requirements.
Summary To solve the present technical problems, a method and a device for automatically generating banner images of a target object in batches are provided in the present invention, to overcome the problems of the long manual designing period, heavy workload, highly repeated work, and low material utilization.
To solve the aforementioned one or more technical problems, the present invention proposes the strategies of that:
from the first perspective, a method for automatically generating banner images of a target object in batches, comprising:
proccssing a batch of target object images, and acquiring a target object main body of each target objcct imagc and targct object image information;
matching at lcast two target master sets for each target objcct main body according to the target objcct main body and the target objcct image information from a mastcr set library;
acquiring laycr information of each target mastcr set and an elcmcnt corrcsponding to each laycr of each target mastcr set according to the laycr information, and determining a matching color of texts for each layer according to layers of each grcyscalc targct mastcr set; and generating a batch of banner imagcs according to the target objcct main body, the targct object imagc information, at least two targct master scts, the elements for individual layers, thc texts, and the text matching color.
Furthermore, the described processing of a batch of targct objcct images, and acquisition of a target objcct main body of each target objcct image and targct object image information includes:
segmenting imagcs for each targct object by image scgmcntation, smoothing scgmcntation results to acquire a target object main body for each targct object image; and acquiring target object image information for each target image according to each described target object image and the described target objcct main body.
Furthermore, the described process of matching at lcast two targct master scts for each targct object main body according to the targct object main body and the target object imagc information from a master set library includcs:
calculating distances between each described targct object main body and cach master set in the master set library according to the dcscribcd target object imagc information, and sclecting at lcast two master sets with the shortest distance from the described target object main body as the target mastcr set.
Furthermorc, the described process of acquiring laycr information of each targct mastcr set, acquiring, an element corresponding to each layer of cach target mastcr set according to the layer information, and determining a matching color of texts for each layer according to layers of each grayscale targct mastcr set include:
structurally analyzing cach of the &scribed at lcast two target mastcr scts to acquire laycr information of each targct mastcr set;
selecting cicmcnts matching each layer of each target master set from the material pool according to thc &scribed layer information of each target mastcr set; and converting layers of each mastcr set into grcyscale, and determining the described text matching color for each laycr according to the color of the grcyscalc laycr.
TARGET OBJECT IN BATCHES
Technical field The present invention is related to the field of image processing technologies, in particular, to a method and device for automatically generating banner images of a target object in batches.
Background With the rapid development and prevalence of intcrnct technologies, online shopping becomes an important part for more and more people. Each c-commerce platforms decorate the product images and add promotion texts to promote information and attract customers, improving commodity transaction rates.
Therefore, generating a high-quality banner image plays a significant role.
Before image processing technologies were applied to the field of visual design, the banner generation is maj orly completed by designer. In the process, designs spend a great amount of time on simple tasks, such as text modifications and spot expansions, wherein the massive redundant work brings limited benefits to designers' skill development. In the meanwhile, precise marketing is the further demand for e-commerce. With the background of massive volume, the poster spots on the home page are required to display in adherence with the principle of "thousands of people-recommendation" (wherein "thousands of people-recommendation" is a recommendation algorithm that can construct customer preference models based on the massive database of e-commerce websites to promote products according to features matching with customer preferences; and the contents displayed on the website read by target customers, to identify potential buyers for sellers for precision marketing). Hereby, high poster generation efficiency is also required.
Therefore, a new banner image generation method is demanded to address the aforementioned requirements.
Summary To solve the present technical problems, a method and a device for automatically generating banner images of a target object in batches are provided in the present invention, to overcome the problems of the long manual designing period, heavy workload, highly repeated work, and low material utilization.
To solve the aforementioned one or more technical problems, the present invention proposes the strategies of that:
from the first perspective, a method for automatically generating banner images of a target object in batches, comprising:
proccssing a batch of target object images, and acquiring a target object main body of each target objcct imagc and targct object image information;
matching at lcast two target master sets for each target objcct main body according to the target objcct main body and the target objcct image information from a mastcr set library;
acquiring laycr information of each target mastcr set and an elcmcnt corrcsponding to each laycr of each target mastcr set according to the laycr information, and determining a matching color of texts for each layer according to layers of each grcyscalc targct mastcr set; and generating a batch of banner imagcs according to the target objcct main body, the targct object imagc information, at least two targct master scts, the elements for individual layers, thc texts, and the text matching color.
Furthermore, the described processing of a batch of targct objcct images, and acquisition of a target objcct main body of each target objcct image and targct object image information includes:
segmenting imagcs for each targct object by image scgmcntation, smoothing scgmcntation results to acquire a target object main body for each targct object image; and acquiring target object image information for each target image according to each described target object image and the described target objcct main body.
Furthermore, the described process of matching at lcast two targct master scts for each targct object main body according to the targct object main body and the target object imagc information from a master set library includcs:
calculating distances between each described targct object main body and cach master set in the master set library according to the dcscribcd target object imagc information, and sclecting at lcast two master sets with the shortest distance from the described target object main body as the target mastcr set.
Furthermorc, the described process of acquiring laycr information of each targct mastcr set, acquiring, an element corresponding to each layer of cach target mastcr set according to the layer information, and determining a matching color of texts for each layer according to layers of each grayscale targct mastcr set include:
structurally analyzing cach of the &scribed at lcast two target mastcr scts to acquire laycr information of each targct mastcr set;
selecting cicmcnts matching each layer of each target master set from the material pool according to thc &scribed layer information of each target mastcr set; and converting layers of each mastcr set into grcyscale, and determining the described text matching color for each laycr according to the color of the grcyscalc laycr.
2 Furthermore, after matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library, the described method further includes:
judging that if thc sizes of the described at least two target master sets are identical to the target size; and whcrc if the sizes arc not identical, expanding the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size.
Furthermore, the described method further includes:
extracting the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and selecting banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner image scores.
From the other perspective, an automatic banner image batch generation device for target objects is provided, comprising:
an image processing module, configured to process a batch of target object images, and acquire a target object main body of each target object image and targct object image information;
a master sct matching module, configured to match at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library;
a material acquisition module, configured to acquire layer information of each target master set and an clement corresponding to each layer of each target master set according to the layer information, and determine a matching color of texts for each layer according to layers of each grcyscalc target master set;
and a banner image generation module, configured to generate a batch of banner images according to the target object main body, the target object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
Furthermore, the described image processing module comprises:
an image segmentation unit, configured to segment images for each target object by image segmentation;
an image optimization unit, configured to smooth segmentation results to acquire a target object main body for each target object image; and an information extracting unit, configured to acquire target object image information for each target image according to each described target object image and the described target object main body.
Furthermore, the described master set matching module comprises:
judging that if thc sizes of the described at least two target master sets are identical to the target size; and whcrc if the sizes arc not identical, expanding the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size.
Furthermore, the described method further includes:
extracting the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and selecting banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner image scores.
From the other perspective, an automatic banner image batch generation device for target objects is provided, comprising:
an image processing module, configured to process a batch of target object images, and acquire a target object main body of each target object image and targct object image information;
a master sct matching module, configured to match at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library;
a material acquisition module, configured to acquire layer information of each target master set and an clement corresponding to each layer of each target master set according to the layer information, and determine a matching color of texts for each layer according to layers of each grcyscalc target master set;
and a banner image generation module, configured to generate a batch of banner images according to the target object main body, the target object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
Furthermore, the described image processing module comprises:
an image segmentation unit, configured to segment images for each target object by image segmentation;
an image optimization unit, configured to smooth segmentation results to acquire a target object main body for each target object image; and an information extracting unit, configured to acquire target object image information for each target image according to each described target object image and the described target object main body.
Furthermore, the described master set matching module comprises:
3 a distance calculation unit, configurcd to calculate distances between each described target object main body and each mastcr sct in the mastcr sct library according to the described target object image information; and a master sct selection unit, configurcd to select at least two master scts with the shortest distance from the described target object main body as thc target master set.
Furthermorc, the described material acquisition module comprises:
a master sct analysis unit, configured to structurally analyzc cach of the describcd at lcast two target master sets to acquire laycr information of each target mastcr sct;
an element matching unit, configured to select elements matching each layer of each target master set from the material pool according to the dcscribed layer information of cach target master sct; and a text matching color unit, configurcd to process layers of each mastcr sct into greyscalc, and dcterminc the &scribed text matching color for cach laycr according to thc color of the greyscale laycr.
Furthermore, the described device further includes:
a size judgement modulc, configured to judge that if the sizes of thc described at least two target master sets are identical to the target size; and a sizc expansion modulc, configured to expand the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets to be the same as the target size what if thc sizes arc not identical.
Furthermore, the described device further includes:
a banner imagc evaluating module, configurcd to extract the target features of each banner image in thc describcd batch of banncr imagcs to calculate a scorc for cach banncr imagc bascd on all target features; and a banner imagc selection module, configured to select banner images with scorcs satisfying prc-set conditions to be outputted from the described batch of banncr imagcs according to the described banner image scores.
The benefits provided by the technical proposal of the present invention include that:
1. thc mcthod and device for automatically gcncrating banncr images of a target object in batches provided in the prcscnt invention first matches multiple master scts from a pre-maintained mastcr sct library according to the batch of target object imagcs uploaded by users, dctermines elements for laycrs and text matching color according to the mastcr sets, and generates a batch of banncr imagcs according to the batch of target object images, multiplc mastcr scts, elements for layers, and text matching color, wherein the design period is reduced to prevent rcdundant work and improve banncr image generating efficiency;
2. thc mcthod and device for automatically gcncrating banncr images of a target object in batches providcd in the present invention segments thc batch of target object imagcs by imagc segmentation
Furthermorc, the described material acquisition module comprises:
a master sct analysis unit, configured to structurally analyzc cach of the describcd at lcast two target master sets to acquire laycr information of each target mastcr sct;
an element matching unit, configured to select elements matching each layer of each target master set from the material pool according to the dcscribed layer information of cach target master sct; and a text matching color unit, configurcd to process layers of each mastcr sct into greyscalc, and dcterminc the &scribed text matching color for cach laycr according to thc color of the greyscale laycr.
Furthermore, the described device further includes:
a size judgement modulc, configured to judge that if the sizes of thc described at least two target master sets are identical to the target size; and a sizc expansion modulc, configured to expand the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets to be the same as the target size what if thc sizes arc not identical.
Furthermore, the described device further includes:
a banner imagc evaluating module, configurcd to extract the target features of each banner image in thc describcd batch of banncr imagcs to calculate a scorc for cach banncr imagc bascd on all target features; and a banner imagc selection module, configured to select banner images with scorcs satisfying prc-set conditions to be outputted from the described batch of banncr imagcs according to the described banner image scores.
The benefits provided by the technical proposal of the present invention include that:
1. thc mcthod and device for automatically gcncrating banncr images of a target object in batches provided in the prcscnt invention first matches multiple master scts from a pre-maintained mastcr sct library according to the batch of target object imagcs uploaded by users, dctermines elements for laycrs and text matching color according to the mastcr sets, and generates a batch of banncr imagcs according to the batch of target object images, multiplc mastcr scts, elements for layers, and text matching color, wherein the design period is reduced to prevent rcdundant work and improve banncr image generating efficiency;
2. thc mcthod and device for automatically gcncrating banncr images of a target object in batches providcd in the present invention segments thc batch of target object imagcs by imagc segmentation
4 algorithm, smooths the segmentation results, and acquires target object main bodies, wherein the image quality of the target object main bodies is improved to improve finalized banner image quality; and 3. the method and device for automatically generating banner images of a target object in batches provided in the present invention expand the size of master sets to ensure that the master sets have identical sizes with the size desired by users, wherein only a small amount of master sets maintained in the master set library can cover all sizes to improve master set utilization.
Brief descriptions of the drawings For a better explanation of the technical proposal of embodiments in the present invention, the accompanying drawings are briefly introduced in the following. Obviously, the following drawings represent only a portion of embodiments of the present invention. Those skilled in the art are able to create other drawings according to the accompanying drawings without making creative efforts.
Fig. 1 is a flow diagram of the method for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention;
Fig. 2 is a flow diagram for the process of processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information in an illustrative embodiment of the present invention;
Fig. 3 is a flow diagram for the process of acquiring layer information of each target master set and an clement corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each greyscale target master set in an illustrative embodiment of the present invention; and Fig. 4 is a structure diagram of the device for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention.
Detail descriptions In order to make the objective, the technical scheme, and the advantages of the present invention clearer, the present invention will be explained further in detail precisely below with references to the accompanying drawings. Obviously, the embodiments described below are only a portion of embodiments of the present invention and cannot represent all possible embodiments. Based on the embodiments in the present invention, the other applications by those skilled in the art without any creative works are falling within the scope of the present invention.
Fig. 1 is a flow diagram of the method for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention. As shown in Fig. 1, the described method comprises the following procedures:
S I, processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information.
In detail, in the present invention embodiments, the target object main bodies include commodities.
The target object main bodies are the major object to be emphasized by the banner images. Generating a good banner image first requires a high-quality target object image (as the target object main body).
Therefore, the batch of target object images uploaded by users should be processed in advance to acquire a target object main body of each target object image and target object image information. In particular, the target object image information includes color of the target object main body, category, applicable scenarios, styles, sizes, and other information.
To clarify, to further improve banner image quality, in the present embodiment, before processing the batch of target object images, the pre-processing of the batch of target object images is further required.
The pre-processing at least includes quality judgement for the batch of target object images uploaded by users to filter out low-resolution blurry images.
S2, matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library.
In detail, in order to generate banner images satisfying structural aesthetics, the pre-maintained master set library is included in the present invention. The master sets in the master set library are maintained according to certain dimensions, such as size, style, category, color, scenario, layout, etc. In other words, master sets with different scenarios, sizes, and layouts are maintained in the master set library.
When banner images are generated, each banner image corresponds to a master set. Wherein a batch of banner images are generated to be selected by users while the banner image generation is diversified, in the present embodiment, at least two target master sets are selected to match with each target object main body according to the target object main body and the target object image information from a master set library.
Banner images generated from each type of target master set compose a series, wherein the generated banner images satisfies aesthetics requirements and are diversified.
53, acquiring layer information of each target master set and an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each greyscale target master set.
In detail, each master set in the master set library includes structural layers with standard names.
As an example, the master sets in the present embodiment include multiple layers, such as the target object main body layer, text layer, and background layer. To clarify, in the present embodiment, the number of layers in master sets is not limited and can be defined according to practical use by users. After the master set is matched, layer information of each target master set is acquired.
Elements corresponding to each layer are obtained based on layer information. Then the text matching colors for layers are determined.
S4, generating a batch of banner images according to the target object main body, the targct object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
In detail, the acquired target object main bodies, target object image information, target master sets, elements for each layer, texts, and text matching colors from the aforementioned steps are combined to generate a batch of banner images. As a preferred embodiment, in the present invention, the batch generation of banner images is abstracted to be a process of feature matching.
In practice, in the stage of feature extraction of the matching algorithm, the layers in selected master sets are abstracted into different dimensions, such as color, size, theme, shape, texture, space, ctc, wherein the dimensions are quantified as feature templates. For the feature extraction, corresponding elements in each layer are matched from the material library, wherein the feature distance between elements and templates are calculated to determine that whether the elements and templates are matching. Then the banner images are generated according to the matching results. To clarify, in the present embodiment, the particular matching algorithm is not restricted, wherein users can select and set according to practical demands.
Fig. 2 is a flow diagram for the process of processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information in an illustrative embodiment of the present invention. As shown in Fig. 2, as a preferred embodiment, the described processing of a batch of target object images, and acquisition of a target object main body of each target object image and target object image information includes:
S101, segmenting images for each target object by image segmentation, smoothing segmentation results to acquire a target object main body for each target object image.
In detail, the image segmentation algorithm based on deep learning can be used to segment each image of the batch of target object images. in other words, the target object main bodies are preliminarily segmented from the background, to separate the target object main bodies from the background. Then, the anti-Aliasing sampling algorithm is applied to the segmented results (i.e. the preliminary segmented target object main bodies) for the edge smoothing process, wherein the high-quality target object main bodies are exported. As an example, the export results can be transparent prints of the target object main bodies.
S102, acquiring target object image information for each target image according to each described target object image and the described target object main body. In detail, the convolutional neural network (CNN) can be applied to extract target object image information for each target object image, wherein the target object image information includes the color of the target object main body, category, applicable scenarios, styles, sizes, and other information.
As a preferred embodiment, in the present embodiment, thc described process of matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library includes:
calculating distances between each described target object main body and each master set in the master set library according to the described target object image information, and selecting at least two master sets with the shortest distance from the described target object main body as the target master set.
In detail, in the present embodiment, the Euclidean distance is used to search for the closest master set as the design drafts for banner images. To clarify, the master sets obtained each time are not identical.
By setting a distance threshold, multiple target master sets can be selected, and banner images generated from each type of target master set compose a series, wherein the generated banner images satisfies aesthetics requirements and are diversified.
Fig. 3 is a flow diagram for the process of acquiring layer information of each target master set and an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each grcyscalc target master set in an illustrative embodiment of the present invention. As shown in Fig. 3, as a preferred embodiment, the described process of acquiring layer information of each target master set, acquiring, an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each grayscalc target master set include:
S301, structurally analyzing each of the described at least two target master sets to acquire layer information of each target master set.
In detail, after the target master sets are matched, each target master set is structurally analyzed to acquire layer information corresponding to each target master set. The layer information includes background, logo, texts, decorations, commodity area decorations, text area decorations, discrete decorations, edge decorations, outline decorations, etc.
S302, selecting elements matching each layer of each target master set from the material pool according to the described layer information of each target master set.
In detail, in the present embodiment, a material pool is constructed according to different layer information of the master sets, wherein the material pool collects elements corresponding to each layer. the elements are updated to the material pool via the wavelet decomposition for label maintenance, such as aspect ratio, color, style, shape, texture, space, size, applicable scenarios, etc. A single clement can correspond to multiple dimension labels. The layer information for each target master set is acquired according to the aforementioned steps, and the elements corresponding to each layer of each master set are selected from the material pool according to the matching rule algorithm. To clarify, in the present embodiment, the particular matching rule algorithms are not restricted and users can select and set according to practical demands.
As a preferred embodiment, in the present embodiment, the material pool further maintains the background image in advance. By investigating and counting service provider banner sizes, the maximum width or height used by the service provider is not larger than 1246. To adapt size requirements for different services, in the present embodiment, all backgrounds are maintained as 1300 x1300 in the material pool, and respectively compressed into the maximum length required by users during image construction. Then the user required size is cropped according to the center of the compressed background image. Therefore, the image construction only involves compression and cropping, to prevent image distortion by stretching and ensuring image quality. To clarify, the aforementioned background image size is an illustrative example, and shall not restrict the present embodiment. In practice, users can maintain the background images in the material pool according to practical needs.
S303, converting layers of each master set into greyscale, and determining the described text matching color for each layer according to the color of the greyscalc layer.
In detail, the banner image texts can be uploaded by users. The text matching color can be determined based on the color shades of the text layer. In practice, each layer of the target master set is converted in greyscalc, then the color shades are determined according to a certain threshold to determine text matching color based on the color shades, wherein text matching colors can be white or black, to prevent non-recognizable texts due to low color contrast.
As a preferred embodiment, in the present embodiment, after matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library, the described method further includes:
judging that if the sizes of the described at least two target master sets are identical to the target size; and where if the sizes are not identical, expanding the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size.
In detail, due to diverse user demands, master sets in the master set library cannot cover all possible user input sizes. When the user input size is different from an acquired master set, the master set is expanded according to user demand size (as the banner image size defined by the user) based on the principle of the fixed relative location of layer elements to the banner image, so as to obtain the master set expanded to the size required by the user. In other words, the topological relation of the relative location of the layer elements to the banner image with the global absolute locations is identified as the target function for expansion. The expansion range is 50% of the aspect ratio. Therefore, only a small number of master sets arc maintained in the master set library to cover all possible sizes, ensuring master set diversity. In practice, each master set layout is maintained for four different ratios to cover all sizes on the platform.
As a preferred embodiment, in the present embodiment, the described method further includes:
extracting the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and selecting banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner image scores.
In detail, to provide better banners to be selected by users, in the present embodiment, the generated banner images are imported into thc banner evaluation system to bc evaluated.
Thc batch of banner images arc sorted based on scores (as the evaluation results) and a batch of finalized banner images arc selected according to the sorting result, wherein the finalized banner images are provided to be selected by users.
The banner evaluation system majorly evaluates banncr images according to the evaluating objects and evaluating dimensions. In the present embodiment, the target features extracted from the banner images are identified as the evaluating objects for scoring banner images. The evaluating objects are elements to be evaluated in banner images. Currently, the main evaluating objects include harmonic color blending, proper layouts, uniformed styles, harmonic commodity images, etc. The evaluating dimensions imply that banners generated from different dimensions are evaluated. With considering all factors such as design and online performance, in the present embodiment, the evaluating dimensions are combined by evaluations from the designers, exposure click ratios, and general offline users.
Massive numbers of design works are assigned as training data to construct the banner evaluation system. In practice, in the training of the banner evaluation model, the cores are determined by operating personnel and designers then imported into the network for training and result exporting. In the model validation stage, the online exposure click ratios in the exported results are used to determine corresponding banner score labels. The banner images are then displayed according to the scores and returned to the model as feedbacks to reinforce the learning, wherein the evaluation mechanism is refined from multiple dimensions.
To clarify, in the present embodiment, pre-set conditions arc set for selecting banner images satisfying the pre-set conditions from the batch of the banner image to be selected by users.
Fig. 4 is a structure diagram of the device for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention. As shown in Fig. 4, the described device comprises:
an image processing module, configured to process a batch of target object images, and acquire a target object main body of each target object image and target object image information;
a master set matching module, configured to match at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library;
a material acquisition module, configured to acquirc layer information of cach target master set and an clement corresponding to each layer of each target master set according to the layer information, and determine a matching color of texts for each layer according to layers of each greyscale target master set;
and a banner image generation module, configured to generate a batch of banner images according to the target object main body, the target object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
As a preferred embodiment, in the present embodiment, the described image processing module comprises:
an image segmentation unit, configured to segment images for each target object by image segmentation;
an image optimization unit, configured to smooth segmentation results to acquire a target object main body for each target object image; and an information extracting unit, configured to acquire target object image information for each target image according to each described target object image and the described target object main body.
As a preferred embodiment, in the present embodiment, the described master set matching module comprises:
a distance calculation unit, configured to calculate distances between each described target object main body and each master set in the master set library according to the described target object image information; and a master set selection unit, configured to select at least two master sets with the shortest distance from the described target object main body as the target master set.
As a preferred embodiment, in the present embodiment, the described material acquisition module comprises:
a master set analysis unit, configured to structurally analyze each of the described at least two target master sets to acquire layer information of each target master set;
an element matching unit, configured to select elements matching each layer of each target master set from the material pool according to the described layer information of each target master set; and a text matching color unit, configured to process layers of each master set into greyscale, and determine the described text matching color for each layer according to the color of the greyscale layer.
As a preferred embodiment, in the present embodiment, the described device further includes:
a size judgement module, configured to judge that if the sizes of the described at least two target master sets are identical to the target size; and a size expansion module, configured to expand the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size where if the sizes are not identical.
As a prcferred embodiment, in the prescnt embodiment, the described dcvicc further includes:
a banner imagc evaluating module, configured to extract the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and a banner imagc selection module, configured to select banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner imagc scores.
In conclusion, the benefits provided by the technical proposal of the present invention include that:
1. the method and device for automatically generating banner images of a target object in batches provided in the present invention first matches multiple master sets from a pre-maintained master set library according to the batch of target object images uploaded by users, determines elements for layers and text matching color according to the master sets, and generates a batch of banner images according to the batch of target object images, multiple master sets, clements for layers, and text matching color, wherein the design period is reduced to prevent redundant work and improve banner imagc generating efficiency;
2. the method and device for automatically generating banner images of a target object in batches provided in the present invention segments the batch of target object images by imagc segmentation algorithm, smooths the segmentation results, and acquires target object main bodies, wherein the image quality of the targct object main bodics is improved to improve finalized banner image quality; and 3. the mcthod and device for automatically generating banncr images of a target object in batches provided in the present invention expand the size of master sets to ensure that the master sets have identical sizes with the size desired by users, wherein only a small amount of master sets maintained in the master set library can cover all sizes to improve master set utilization.
To clarify, whcn the automatic banner imagc batch generation service is invoked in the automatic banner image batch gencration device in the aforementioned embodiments, the described functional module configurations are used for illustration only. In practical applications, the described functions can be assigned to different functional modules according to practical demands, wherein the internal structural configuration of the device is divided into different functional modules to perform all or a portion of the described functions. Besides, the aforementioned automatic banner image batch generation device in the embodiment adopts the same concepts in the described automatic banner image batch generation method embodiments. The described device is based on the implementation of the automatic banner image batch generation method, whereas the detailed procedures can be referred to the method embodiments and are not explained in further detail.
All or portions of the aforementioned procedures are comprehensible for those skilled in the art, and may be achieved by the computer program configured for sending commands to the related hardware.
The computer programs can be stored in computer readable memory units, wherein procedures of the aforementioned automatic banner image batch generation method are performed when the described computer programs are executed. The storage medium may include ROM/RAM, diskettes, disc, memory cards, etc.
The aforementioned contents of preferred embodiments in the present invention shall not limit applications of the present invention. Therefore, all alterations, modifications, equivalence, improvements of the present invention fall within the scope of the present invention.
Brief descriptions of the drawings For a better explanation of the technical proposal of embodiments in the present invention, the accompanying drawings are briefly introduced in the following. Obviously, the following drawings represent only a portion of embodiments of the present invention. Those skilled in the art are able to create other drawings according to the accompanying drawings without making creative efforts.
Fig. 1 is a flow diagram of the method for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention;
Fig. 2 is a flow diagram for the process of processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information in an illustrative embodiment of the present invention;
Fig. 3 is a flow diagram for the process of acquiring layer information of each target master set and an clement corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each greyscale target master set in an illustrative embodiment of the present invention; and Fig. 4 is a structure diagram of the device for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention.
Detail descriptions In order to make the objective, the technical scheme, and the advantages of the present invention clearer, the present invention will be explained further in detail precisely below with references to the accompanying drawings. Obviously, the embodiments described below are only a portion of embodiments of the present invention and cannot represent all possible embodiments. Based on the embodiments in the present invention, the other applications by those skilled in the art without any creative works are falling within the scope of the present invention.
Fig. 1 is a flow diagram of the method for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention. As shown in Fig. 1, the described method comprises the following procedures:
S I, processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information.
In detail, in the present invention embodiments, the target object main bodies include commodities.
The target object main bodies are the major object to be emphasized by the banner images. Generating a good banner image first requires a high-quality target object image (as the target object main body).
Therefore, the batch of target object images uploaded by users should be processed in advance to acquire a target object main body of each target object image and target object image information. In particular, the target object image information includes color of the target object main body, category, applicable scenarios, styles, sizes, and other information.
To clarify, to further improve banner image quality, in the present embodiment, before processing the batch of target object images, the pre-processing of the batch of target object images is further required.
The pre-processing at least includes quality judgement for the batch of target object images uploaded by users to filter out low-resolution blurry images.
S2, matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library.
In detail, in order to generate banner images satisfying structural aesthetics, the pre-maintained master set library is included in the present invention. The master sets in the master set library are maintained according to certain dimensions, such as size, style, category, color, scenario, layout, etc. In other words, master sets with different scenarios, sizes, and layouts are maintained in the master set library.
When banner images are generated, each banner image corresponds to a master set. Wherein a batch of banner images are generated to be selected by users while the banner image generation is diversified, in the present embodiment, at least two target master sets are selected to match with each target object main body according to the target object main body and the target object image information from a master set library.
Banner images generated from each type of target master set compose a series, wherein the generated banner images satisfies aesthetics requirements and are diversified.
53, acquiring layer information of each target master set and an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each greyscale target master set.
In detail, each master set in the master set library includes structural layers with standard names.
As an example, the master sets in the present embodiment include multiple layers, such as the target object main body layer, text layer, and background layer. To clarify, in the present embodiment, the number of layers in master sets is not limited and can be defined according to practical use by users. After the master set is matched, layer information of each target master set is acquired.
Elements corresponding to each layer are obtained based on layer information. Then the text matching colors for layers are determined.
S4, generating a batch of banner images according to the target object main body, the targct object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
In detail, the acquired target object main bodies, target object image information, target master sets, elements for each layer, texts, and text matching colors from the aforementioned steps are combined to generate a batch of banner images. As a preferred embodiment, in the present invention, the batch generation of banner images is abstracted to be a process of feature matching.
In practice, in the stage of feature extraction of the matching algorithm, the layers in selected master sets are abstracted into different dimensions, such as color, size, theme, shape, texture, space, ctc, wherein the dimensions are quantified as feature templates. For the feature extraction, corresponding elements in each layer are matched from the material library, wherein the feature distance between elements and templates are calculated to determine that whether the elements and templates are matching. Then the banner images are generated according to the matching results. To clarify, in the present embodiment, the particular matching algorithm is not restricted, wherein users can select and set according to practical demands.
Fig. 2 is a flow diagram for the process of processing a batch of target object images, and acquiring a target object main body of each target object image and target object image information in an illustrative embodiment of the present invention. As shown in Fig. 2, as a preferred embodiment, the described processing of a batch of target object images, and acquisition of a target object main body of each target object image and target object image information includes:
S101, segmenting images for each target object by image segmentation, smoothing segmentation results to acquire a target object main body for each target object image.
In detail, the image segmentation algorithm based on deep learning can be used to segment each image of the batch of target object images. in other words, the target object main bodies are preliminarily segmented from the background, to separate the target object main bodies from the background. Then, the anti-Aliasing sampling algorithm is applied to the segmented results (i.e. the preliminary segmented target object main bodies) for the edge smoothing process, wherein the high-quality target object main bodies are exported. As an example, the export results can be transparent prints of the target object main bodies.
S102, acquiring target object image information for each target image according to each described target object image and the described target object main body. In detail, the convolutional neural network (CNN) can be applied to extract target object image information for each target object image, wherein the target object image information includes the color of the target object main body, category, applicable scenarios, styles, sizes, and other information.
As a preferred embodiment, in the present embodiment, thc described process of matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library includes:
calculating distances between each described target object main body and each master set in the master set library according to the described target object image information, and selecting at least two master sets with the shortest distance from the described target object main body as the target master set.
In detail, in the present embodiment, the Euclidean distance is used to search for the closest master set as the design drafts for banner images. To clarify, the master sets obtained each time are not identical.
By setting a distance threshold, multiple target master sets can be selected, and banner images generated from each type of target master set compose a series, wherein the generated banner images satisfies aesthetics requirements and are diversified.
Fig. 3 is a flow diagram for the process of acquiring layer information of each target master set and an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each grcyscalc target master set in an illustrative embodiment of the present invention. As shown in Fig. 3, as a preferred embodiment, the described process of acquiring layer information of each target master set, acquiring, an element corresponding to each layer of each target master set according to the layer information, and determining a matching color of texts for each layer according to layers of each grayscalc target master set include:
S301, structurally analyzing each of the described at least two target master sets to acquire layer information of each target master set.
In detail, after the target master sets are matched, each target master set is structurally analyzed to acquire layer information corresponding to each target master set. The layer information includes background, logo, texts, decorations, commodity area decorations, text area decorations, discrete decorations, edge decorations, outline decorations, etc.
S302, selecting elements matching each layer of each target master set from the material pool according to the described layer information of each target master set.
In detail, in the present embodiment, a material pool is constructed according to different layer information of the master sets, wherein the material pool collects elements corresponding to each layer. the elements are updated to the material pool via the wavelet decomposition for label maintenance, such as aspect ratio, color, style, shape, texture, space, size, applicable scenarios, etc. A single clement can correspond to multiple dimension labels. The layer information for each target master set is acquired according to the aforementioned steps, and the elements corresponding to each layer of each master set are selected from the material pool according to the matching rule algorithm. To clarify, in the present embodiment, the particular matching rule algorithms are not restricted and users can select and set according to practical demands.
As a preferred embodiment, in the present embodiment, the material pool further maintains the background image in advance. By investigating and counting service provider banner sizes, the maximum width or height used by the service provider is not larger than 1246. To adapt size requirements for different services, in the present embodiment, all backgrounds are maintained as 1300 x1300 in the material pool, and respectively compressed into the maximum length required by users during image construction. Then the user required size is cropped according to the center of the compressed background image. Therefore, the image construction only involves compression and cropping, to prevent image distortion by stretching and ensuring image quality. To clarify, the aforementioned background image size is an illustrative example, and shall not restrict the present embodiment. In practice, users can maintain the background images in the material pool according to practical needs.
S303, converting layers of each master set into greyscale, and determining the described text matching color for each layer according to the color of the greyscalc layer.
In detail, the banner image texts can be uploaded by users. The text matching color can be determined based on the color shades of the text layer. In practice, each layer of the target master set is converted in greyscalc, then the color shades are determined according to a certain threshold to determine text matching color based on the color shades, wherein text matching colors can be white or black, to prevent non-recognizable texts due to low color contrast.
As a preferred embodiment, in the present embodiment, after matching at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library, the described method further includes:
judging that if the sizes of the described at least two target master sets are identical to the target size; and where if the sizes are not identical, expanding the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size.
In detail, due to diverse user demands, master sets in the master set library cannot cover all possible user input sizes. When the user input size is different from an acquired master set, the master set is expanded according to user demand size (as the banner image size defined by the user) based on the principle of the fixed relative location of layer elements to the banner image, so as to obtain the master set expanded to the size required by the user. In other words, the topological relation of the relative location of the layer elements to the banner image with the global absolute locations is identified as the target function for expansion. The expansion range is 50% of the aspect ratio. Therefore, only a small number of master sets arc maintained in the master set library to cover all possible sizes, ensuring master set diversity. In practice, each master set layout is maintained for four different ratios to cover all sizes on the platform.
As a preferred embodiment, in the present embodiment, the described method further includes:
extracting the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and selecting banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner image scores.
In detail, to provide better banners to be selected by users, in the present embodiment, the generated banner images are imported into thc banner evaluation system to bc evaluated.
Thc batch of banner images arc sorted based on scores (as the evaluation results) and a batch of finalized banner images arc selected according to the sorting result, wherein the finalized banner images are provided to be selected by users.
The banner evaluation system majorly evaluates banncr images according to the evaluating objects and evaluating dimensions. In the present embodiment, the target features extracted from the banner images are identified as the evaluating objects for scoring banner images. The evaluating objects are elements to be evaluated in banner images. Currently, the main evaluating objects include harmonic color blending, proper layouts, uniformed styles, harmonic commodity images, etc. The evaluating dimensions imply that banners generated from different dimensions are evaluated. With considering all factors such as design and online performance, in the present embodiment, the evaluating dimensions are combined by evaluations from the designers, exposure click ratios, and general offline users.
Massive numbers of design works are assigned as training data to construct the banner evaluation system. In practice, in the training of the banner evaluation model, the cores are determined by operating personnel and designers then imported into the network for training and result exporting. In the model validation stage, the online exposure click ratios in the exported results are used to determine corresponding banner score labels. The banner images are then displayed according to the scores and returned to the model as feedbacks to reinforce the learning, wherein the evaluation mechanism is refined from multiple dimensions.
To clarify, in the present embodiment, pre-set conditions arc set for selecting banner images satisfying the pre-set conditions from the batch of the banner image to be selected by users.
Fig. 4 is a structure diagram of the device for automatically generating banner images of a target object in batches in an illustrative embodiment of the present invention. As shown in Fig. 4, the described device comprises:
an image processing module, configured to process a batch of target object images, and acquire a target object main body of each target object image and target object image information;
a master set matching module, configured to match at least two target master sets for each target object main body according to the target object main body and the target object image information from a master set library;
a material acquisition module, configured to acquirc layer information of cach target master set and an clement corresponding to each layer of each target master set according to the layer information, and determine a matching color of texts for each layer according to layers of each greyscale target master set;
and a banner image generation module, configured to generate a batch of banner images according to the target object main body, the target object image information, at least two target master sets, the elements for individual layers, the texts, and the text matching color.
As a preferred embodiment, in the present embodiment, the described image processing module comprises:
an image segmentation unit, configured to segment images for each target object by image segmentation;
an image optimization unit, configured to smooth segmentation results to acquire a target object main body for each target object image; and an information extracting unit, configured to acquire target object image information for each target image according to each described target object image and the described target object main body.
As a preferred embodiment, in the present embodiment, the described master set matching module comprises:
a distance calculation unit, configured to calculate distances between each described target object main body and each master set in the master set library according to the described target object image information; and a master set selection unit, configured to select at least two master sets with the shortest distance from the described target object main body as the target master set.
As a preferred embodiment, in the present embodiment, the described material acquisition module comprises:
a master set analysis unit, configured to structurally analyze each of the described at least two target master sets to acquire layer information of each target master set;
an element matching unit, configured to select elements matching each layer of each target master set from the material pool according to the described layer information of each target master set; and a text matching color unit, configured to process layers of each master set into greyscale, and determine the described text matching color for each layer according to the color of the greyscale layer.
As a preferred embodiment, in the present embodiment, the described device further includes:
a size judgement module, configured to judge that if the sizes of the described at least two target master sets are identical to the target size; and a size expansion module, configured to expand the sizes of the described at least two target master sets so as for ensuring the size of the described at least two target master sets the same as the target size where if the sizes are not identical.
As a prcferred embodiment, in the prescnt embodiment, the described dcvicc further includes:
a banner imagc evaluating module, configured to extract the target features of each banner image in the described batch of banner images to calculate a score for each banner image based on all target features; and a banner imagc selection module, configured to select banner images with scores satisfying pre-set conditions to be outputted from the described batch of banner images according to the described banner imagc scores.
In conclusion, the benefits provided by the technical proposal of the present invention include that:
1. the method and device for automatically generating banner images of a target object in batches provided in the present invention first matches multiple master sets from a pre-maintained master set library according to the batch of target object images uploaded by users, determines elements for layers and text matching color according to the master sets, and generates a batch of banner images according to the batch of target object images, multiple master sets, clements for layers, and text matching color, wherein the design period is reduced to prevent redundant work and improve banner imagc generating efficiency;
2. the method and device for automatically generating banner images of a target object in batches provided in the present invention segments the batch of target object images by imagc segmentation algorithm, smooths the segmentation results, and acquires target object main bodies, wherein the image quality of the targct object main bodics is improved to improve finalized banner image quality; and 3. the mcthod and device for automatically generating banncr images of a target object in batches provided in the present invention expand the size of master sets to ensure that the master sets have identical sizes with the size desired by users, wherein only a small amount of master sets maintained in the master set library can cover all sizes to improve master set utilization.
To clarify, whcn the automatic banner imagc batch generation service is invoked in the automatic banner image batch gencration device in the aforementioned embodiments, the described functional module configurations are used for illustration only. In practical applications, the described functions can be assigned to different functional modules according to practical demands, wherein the internal structural configuration of the device is divided into different functional modules to perform all or a portion of the described functions. Besides, the aforementioned automatic banner image batch generation device in the embodiment adopts the same concepts in the described automatic banner image batch generation method embodiments. The described device is based on the implementation of the automatic banner image batch generation method, whereas the detailed procedures can be referred to the method embodiments and are not explained in further detail.
All or portions of the aforementioned procedures are comprehensible for those skilled in the art, and may be achieved by the computer program configured for sending commands to the related hardware.
The computer programs can be stored in computer readable memory units, wherein procedures of the aforementioned automatic banner image batch generation method are performed when the described computer programs are executed. The storage medium may include ROM/RAM, diskettes, disc, memory cards, etc.
The aforementioned contents of preferred embodiments in the present invention shall not limit applications of the present invention. Therefore, all alterations, modifications, equivalence, improvements of the present invention fall within the scope of the present invention.
Claims (10)
1. A mcthod for automatically gcnerating banncr images of a targct objcct in batches, comprises:
processing a batch of target object images, and acquiring a target object main body of each target objcct image and target objcct image information;
matching at least two targct master scts for each target objcct main body according to thc target objcct main body and thc targct object image information from a mastcr sct library;
acquiring laycr information of cach targct mastcr sct and an cicmcnt corrcsponding to cach layer of each target master set according to thc laycr information, and dctermining a matching color of tcxts for cach laycr according to laycrs of cach grcyscalc targct mastcr sct;
and gcncrating a batch of banner imagcs according to the targct objcct main body, thc target objcct imagc information, at least two targct master scts, thc cicmcnts for individual laycrs, thc tcxts, and thc tcxt matching color.
processing a batch of target object images, and acquiring a target object main body of each target objcct image and target objcct image information;
matching at least two targct master scts for each target objcct main body according to thc target objcct main body and thc targct object image information from a mastcr sct library;
acquiring laycr information of cach targct mastcr sct and an cicmcnt corrcsponding to cach layer of each target master set according to thc laycr information, and dctermining a matching color of tcxts for cach laycr according to laycrs of cach grcyscalc targct mastcr sct;
and gcncrating a batch of banner imagcs according to the targct objcct main body, thc target objcct imagc information, at least two targct master scts, thc cicmcnts for individual laycrs, thc tcxts, and thc tcxt matching color.
2. Thc automatic banncr imagc batch gcncration mcthod for targct objccts in claim 1, is characterized in that, thc dcscribcd proccssing of a batch of targct objcct imagcs, and acquisition of a targct object main body of each target object imagc and targct object image information includcs:
segmcnting imagcs for cach targct objcct by image scgmentation, smoothing scgmcntation rcsults to acquire a targct object main body for each target object image; and acquiring target objcct imagc information for cach targct imagc according to cach describcd targct objcct image and thc describcd target objcct main body.
segmcnting imagcs for cach targct objcct by image scgmentation, smoothing scgmcntation rcsults to acquire a targct object main body for each target object image; and acquiring target objcct imagc information for cach targct imagc according to cach describcd targct objcct image and thc describcd target objcct main body.
3. Thc automatic banncr imagc batch gcncration mcthod for targct objccts in claim 1 or 2, is characterized in that, the dcscribcd process of matching at least two targct mastcr scts for cach targct objcct main body according to thc targct objcct main body and thc targct objcct imagc information from a mastcr sct library includes:
calculating distances between each describcd target object main body and each master sct in thc master sct library according to thc &scribed targct object imagc information, and sciccting at lcast two mastcr scts with thc shortcst distancc from the &scribed targct object main body as thc targct master set.
calculating distances between each describcd target object main body and each master sct in thc master sct library according to thc &scribed targct object imagc information, and sciccting at lcast two mastcr scts with thc shortcst distancc from the &scribed targct object main body as thc targct master set.
4. The automatic banner imagc batch generation mcthod for targct objccts in claim 1 or 2, is characterized in that, the &scribed process of acquiring layer information of cach targct mastcr sct, acquiring, an elemcnt corrcsponding to cach layer of each target mastcr set according to the layer information, and dctcrmining a matching color of texts for cach layer according to laycrs of each grayscalc target master sct includc:
structurally analyzing each of thc &scribed at least two targct master scts to acquire laycr information of cach targct master sct;
selecting cicmcnts matching each laycr of each target master sct from the material pool according to the described layer information of cach target master sct; and converting laycrs of each master sct into grcyscalc, and detcrmining thc dcscribed text matching color for cach layer according to thc color of thc grcyscalc layer.
structurally analyzing each of thc &scribed at least two targct master scts to acquire laycr information of cach targct master sct;
selecting cicmcnts matching each laycr of each target master sct from the material pool according to the described layer information of cach target master sct; and converting laycrs of each master sct into grcyscalc, and detcrmining thc dcscribed text matching color for cach layer according to thc color of thc grcyscalc layer.
5. The automatic banner imagc batch generation mcthod for target objccts in claim 1 or 2, is characterized in that, after matching at least two target mastcr sets for each target objcct main body according to the target objcct main body and the target objcct image information from a mastcr set library, the dcscribed mcthod furthcr includes:
judging that if thc sizcs of the dcscribed at lcast two target mastcr sets are idcntical to thc targct size; and whcrc if the sizcs arc not idcntical, expanding thc sizcs of thc describcd at lcast two targct mastcr sets so as for cnsuring the sizc of thc describcd at least two targct master scts the same as the targct sizc.
judging that if thc sizcs of the dcscribed at lcast two target mastcr sets are idcntical to thc targct size; and whcrc if the sizcs arc not idcntical, expanding thc sizcs of thc describcd at lcast two targct mastcr sets so as for cnsuring the sizc of thc describcd at least two targct master scts the same as the targct sizc.
6. Thc automatic banncr image batch gcncration mcthod for target objects in claim 1 or 2, is characterized in that, the dcscribcd method furthcr includcs:
cxtracting the targct fcaturcs of cach banner image in the describcd batch of banncr imagcs to calculate a score for cach banner image bascd on all targct features; and selecting banncr imagcs with scorcs satisfying prc-sct conditions to bc outputted from thc dcscribcd batch of banner imagcs according to thc &scribed banner imagc scorcs.
cxtracting the targct fcaturcs of cach banner image in the describcd batch of banncr imagcs to calculate a score for cach banner image bascd on all targct features; and selecting banncr imagcs with scorcs satisfying prc-sct conditions to bc outputted from thc dcscribcd batch of banner imagcs according to thc &scribed banner imagc scorcs.
7. An automatic banncr imagc batch gcncration devicc for targct objccts, compriscs:
an imagc processing module, configured to process a batch of targct objcct images, and acquirc a targct objcct main body of each targct objcct image and targct objcct image information;
a mastcr sct matching module, configurcd to match at lcast two target mastcr scts for cach targct objcct main body according to thc target objcct main body and the targct objcct image information from a master set library;
a material acquisition modulc, configured to acquirc laycr information of each targct mastcr sct and an cicmcnt corrcsponding to each layer of each target mastcr set according to the laycr information, and determine a matching color of texts for cach layer according to laycrs of each grcyscalc target mastcr sct; and a banner image gcneration module, configured to gencrate a batch of banner imagcs according to thc target objcct main body, the targct object image information, at least two target master scts, thc elemcnts for individual layers, the tcxts, and the tcxt matching color.
an imagc processing module, configured to process a batch of targct objcct images, and acquirc a targct objcct main body of each targct objcct image and targct objcct image information;
a mastcr sct matching module, configurcd to match at lcast two target mastcr scts for cach targct objcct main body according to thc target objcct main body and the targct objcct image information from a master set library;
a material acquisition modulc, configured to acquirc laycr information of each targct mastcr sct and an cicmcnt corrcsponding to each layer of each target mastcr set according to the laycr information, and determine a matching color of texts for cach layer according to laycrs of each grcyscalc target mastcr sct; and a banner image gcneration module, configured to gencrate a batch of banner imagcs according to thc target objcct main body, the targct object image information, at least two target master scts, thc elemcnts for individual layers, the tcxts, and the tcxt matching color.
8. The automatic banner imagc batch generation &vice for targct objects in claim 7, is charactcrized in that, thc dcscribcd imagc proccssing modulc compriscs:
an imagc segmentation unit, configured to scgmcnt images for cach target objcct by image segmentation;
an imagc optimization unit, configured to smooth scgmcntation rcsults to acquire a target object main body for cach targct objcct imagc; and an information extracting unit, configured to acquire targct object image information for cach targct imagc according to cach dcscribcd targct objcct imagc and thc dcscribcd targct objcct main body.
an imagc segmentation unit, configured to scgmcnt images for cach target objcct by image segmentation;
an imagc optimization unit, configured to smooth scgmcntation rcsults to acquire a target object main body for cach targct objcct imagc; and an information extracting unit, configured to acquire targct object image information for cach targct imagc according to cach dcscribcd targct objcct imagc and thc dcscribcd targct objcct main body.
9. The automatic banner imagc batch generation dcvice for targct objects in claim 7 or 8, is characterized in that, the dcscribcd mastcr sct matching modulc compriscs:
a distance calculation unit, configured to calculatc distanccs bctwcen cach dcscribcd targct objcct main body and each master set in the master set library according to thc dcscribed targct objcct imagc information; and a mastcr sct scicction unit, configured to sclect at least two master scts with thc shortcst distance from the described targct objcct main body as thc target master set.
a distance calculation unit, configured to calculatc distanccs bctwcen cach dcscribcd targct objcct main body and each master set in the master set library according to thc dcscribed targct objcct imagc information; and a mastcr sct scicction unit, configured to sclect at least two master scts with thc shortcst distance from the described targct objcct main body as thc target master set.
10. Thc automatic banncr imagc batch gcncration dcvicc for targct objects in claim 6 or 7, is characterized in that, thc dcscribcd matcrial acquisition modulc compriscs:
a mastcr sct analysis unit, configurcd to structurally analyzc each of the describcd at lcast two targct mastcr sets to acquire laycr information of each target master sct;
an cicmcnt matching unit, configurcd to select elements matching cach layer of cach targct mastcr set from thc matcrial pool according to the dcscribed layer information of cach target mastcr sct;
and a text matching color unit, configured to process layers of each master sct into grcyscale, and dcterminc the &scribed tcxt matching color for cach laycr according to thc color of thc grcyscalc layer.
a mastcr sct analysis unit, configurcd to structurally analyzc each of the describcd at lcast two targct mastcr sets to acquire laycr information of each target master sct;
an cicmcnt matching unit, configurcd to select elements matching cach layer of cach targct mastcr set from thc matcrial pool according to the dcscribed layer information of cach target mastcr sct;
and a text matching color unit, configured to process layers of each master sct into grcyscale, and dcterminc the &scribed tcxt matching color for cach laycr according to thc color of thc grcyscalc layer.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910777228.4A CN110659371B (en) | 2019-08-21 | 2019-08-21 | Automatic batch generation method and device for banner images of target objects |
| CN201910777228.4 | 2019-08-21 | ||
| PCT/CN2020/096991 WO2021031677A1 (en) | 2019-08-21 | 2020-06-19 | Method and device for automatically generating banner images of target object in batches |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CA3148915A1 true CA3148915A1 (en) | 2021-02-25 |
Family
ID=69037701
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA3148915A Pending CA3148915A1 (en) | 2019-08-21 | 2020-06-19 | Method and device for automatically generating banner images of a target object in batches |
Country Status (3)
| Country | Link |
|---|---|
| CN (1) | CN110659371B (en) |
| CA (1) | CA3148915A1 (en) |
| WO (1) | WO2021031677A1 (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110659371B (en) * | 2019-08-21 | 2022-07-01 | 苏宁云计算有限公司 | Automatic batch generation method and device for banner images of target objects |
| CN111353532A (en) * | 2020-02-26 | 2020-06-30 | 北京三快在线科技有限公司 | Image generation method and device, computer-readable storage medium and electronic device |
| CN113450433A (en) * | 2020-03-26 | 2021-09-28 | 阿里巴巴集团控股有限公司 | Picture generation method and device, computer equipment and medium |
| CN113628105A (en) * | 2020-05-07 | 2021-11-09 | 阿里巴巴集团控股有限公司 | Image processing method, device, storage medium and processor |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2011530737A (en) * | 2008-08-08 | 2011-12-22 | トムソン ライセンシング | Dynamic banner configuration method |
| US11270350B2 (en) * | 2012-11-30 | 2022-03-08 | 3I Avi, Llc | Systems and method for verifying vehicle banner production and image alteration |
| US20140365310A1 (en) * | 2013-06-05 | 2014-12-11 | Machine Perception Technologies, Inc. | Presentation of materials based on low level feature analysis |
| CN107025676B (en) * | 2016-01-25 | 2021-02-02 | 阿里巴巴集团控股有限公司 | Picture template, picture generation method and related device |
| CN108345624A (en) * | 2017-01-24 | 2018-07-31 | 阿里巴巴集团控股有限公司 | Generate the method and device of the page |
| CN108961362A (en) * | 2017-05-27 | 2018-12-07 | 阿里巴巴集团控股有限公司 | A kind of generation method and device of network picture |
| CN109308729B (en) * | 2017-07-27 | 2023-01-24 | 阿里巴巴集团控股有限公司 | Picture synthesis processing method, device and system |
| CN107909631A (en) * | 2017-11-29 | 2018-04-13 | 商派软件有限公司 | A kind of picture synthetic method |
| CN109961493A (en) * | 2017-12-26 | 2019-07-02 | 苏宁云商集团股份有限公司 | Banner Picture Generation Method and device on displayed page |
| CN109741425B (en) * | 2019-01-04 | 2023-07-25 | 广州方硅信息技术有限公司 | Banner picture generation method and device, storage medium and computer equipment |
| CN110659371B (en) * | 2019-08-21 | 2022-07-01 | 苏宁云计算有限公司 | Automatic batch generation method and device for banner images of target objects |
-
2019
- 2019-08-21 CN CN201910777228.4A patent/CN110659371B/en active Active
-
2020
- 2020-06-19 CA CA3148915A patent/CA3148915A1/en active Pending
- 2020-06-19 WO PCT/CN2020/096991 patent/WO2021031677A1/en active Application Filing
Also Published As
| Publication number | Publication date |
|---|---|
| CN110659371B (en) | 2022-07-01 |
| CN110659371A (en) | 2020-01-07 |
| WO2021031677A1 (en) | 2021-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CA3148915A1 (en) | Method and device for automatically generating banner images of a target object in batches | |
| US11636364B2 (en) | Image-based popularity prediction | |
| US20210279642A1 (en) | Mixed-initiative machine learning systems and methods for determining segmentations | |
| CN109308729B (en) | Picture synthesis processing method, device and system | |
| CN111738764B (en) | Method and system for predicting effect of advertisement creativity and generating advertisement creativity | |
| US10699453B2 (en) | Digital media environment for style-aware patching in a digital image | |
| US20110119136A1 (en) | Advertising Forecast and Revenue Systems and Methods | |
| CN111709816A (en) | Service recommendation method, device and equipment based on image recognition and storage medium | |
| CN109543359B (en) | Artificial intelligence package design method and system based on Internet of things big data | |
| CN112053430B (en) | Text-created product design scheme evaluation system adopting mobile terminal and augmented reality technology | |
| CN111178624A (en) | Method for predicting new product demand | |
| CN111243061B (en) | Commodity picture generation method, device and system | |
| CN112214508B (en) | Data processing method and device | |
| WO2021098310A1 (en) | Video generation method and device, and terminal and storage medium | |
| CN117058275A (en) | Commodity propaganda drawing generation method and device, computer equipment and storage medium | |
| CN111061960A (en) | Method for generating user image based on social big data | |
| Jing et al. | Layout Generation for Various Scenarios in Mobile Shopping Applications | |
| CN116246195A (en) | Image-dependent automatic video generation for e-commerce applications | |
| Han et al. | Shapes as product differentiation: Neural network embedding in the analysis of markets for fonts | |
| CN111932315B (en) | Method and device for data display, electronic equipment and computer readable storage medium | |
| CN113297520A (en) | Page design auxiliary processing method and device and electronic equipment | |
| CN113127597A (en) | Processing method and device for search information and electronic equipment | |
| CN113342334B (en) | Visual design releasing method and device | |
| Wang | The interactive embodiment of aesthetics combined with visual object recognition algorithm in graphic design in the field of artificial intelligence | |
| CN113223102A (en) | Image obtaining method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |
|
| EEER | Examination request |
Effective date: 20220916 |