KR102480557B1 - Model generating device for modeling image recommendation based on user input keyword and artificial intelligence - Google Patents

Abstract

A modeling image recommendation model generation device based on a user input keyword and artificial intelligence, according to an embodiment, comprises: a keyword reception unit which receives, from a user, a plurality of main keywords each representing a finished product object classified as a heterogeneous object; a keyword collection unit which collects a plurality of sub-keywords corresponding to each of the plurality of main keywords; an image group reading unit which reads a plurality of groups consisting of images including two-dimensional objects corresponding to part objects indicated by each of the plurality of sub-keywords; an image selection unit which selects one of the images from each of the plurality of groups; a modeling unit which generates a three-dimensional object from the two-dimensional object included in each of the selected image and generates a plurality of modeling images in which each of the three-dimensional objects is combined; and a modeling image management unit which generates an artificial intelligence-based modeling image recommendation model based on the plurality of generated modeling images. Therefore, according to the present invention, an effective combination process can proceed.

Description

Model generating device for modeling image recommendation based on user input keyword and artificial intelligence}

The present invention relates to an apparatus for generating a modeling image recommendation model based on user input keywords and artificial intelligence. An apparatus for generating a modeling image recommendation model based on a user input keyword and artificial intelligence for presenting a method of providing a modeling image corresponding to an object.

A user may search for furniture of interest through a device providing a web service, and may be provided with information on the searched furniture. At this time, the user's search may be performed by inputting various queries related to furniture of interest.

That is, when the user inputs a query related to furniture, the device provides search results related to the query input by the user to the user.

These search results include various data such as text data, music data, image data, and video data. The device obtains and stores search data to provide a search service, and selects and outputs search result information related to furniture from among the search data.

On the other hand, recently, as an extension of convergence technology, there is a need to obtain new search result information with a combination of new keywords that were not thought to be possible to combine in the past. For example, if only keywords related to furniture were input in the past, and keywords related to non-furniture, which is an object other than furniture, were input in a separate process, recently, objects that were considered difficult to combine were input together to create a new object for these objects. There is a growing need for relevance. In addition, there is a need to present a modeling image corresponding to a new type of object by associating this need with a process of inputting a keyword and providing a modeling image resulting therefrom.

The present invention has been derived from the above-described need, and aims to provide users with a modeling image corresponding to a new type of integrated object in which uses are combined by receiving keyword information about objects that are not combined in use. there is

An apparatus for generating a modeling image recommendation model based on a user input keyword and artificial intelligence according to an embodiment includes a keyword receiving unit that receives a plurality of main keywords respectively representing finished product objects classified as heterogeneous objects from a user; a keyword collection unit collecting a plurality of sub-keywords corresponding to each of the plurality of main keywords; an image group reading unit for reading a plurality of groups composed of images including a two-dimensional object corresponding to a part object indicated by each of the plurality of sub-keywords; an image selector selecting one of the images from each of the plurality of groups;

a modeling unit generating 3-dimensional objects from the 2-dimensional objects included in each of the selected images, and generating a plurality of modeling images in which each of the 3-dimensional objects is combined; and

It may include; a modeling image management unit for generating an artificial intelligence-based modeling image recommendation model based on the plurality of generated modeling images.

The image selector may select one of the images from each of the plurality of groups to indicate that each time stamp is similar to each other within a predetermined range and the season of use of each of the finished product objects is similar.

The image selector may select, from each of the plurality of groups, one image indicating that the size of each part of each part object of each tag is similar to each other within a predetermined range.

The modeling unit may create each of the 3D objects based on attribute data of each of the 2D objects.

The modeling image management unit,

The plurality of modeling images are provided to the user, some of the plurality of modeling images are selected and input from the user, and a neural network model is trained using the selected modeling images and the plurality of main keywords. , The artificial intelligence-based modeling image recommendation model may be generated by learning the neural network model.

The modeling image in which each of the 3D objects is combined may represent an integrated object having a new use in a form in which the finished product objects having different uses are combined.

According to the present invention, it is possible to provide a user with a modeling image corresponding to a new type of integrated object in which a purpose is combined by receiving keyword information about objects that are not combined in terms of purpose.

According to the present invention, images of parts corresponding to finished products used in the same season can be combined, and images of unnecessary or awkward combinations can be excluded and provided to the user, so that an effective combination process can proceed.

1 is a block diagram of a modeling image recommendation model generating device 1 based on a user input keyword and artificial intelligence according to an embodiment, and FIG. 2 is a modeling image recommendation model generating device based on a user input keyword and artificial intelligence according to an embodiment. It is a flowchart for explaining the operation of (1), and FIGS. 3 to 5 are referenced to explain FIGS. 1 and 2 .
4 to 5 are perspective views of an actual finished product object and a part object, respectively.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The detailed description of the present invention which follows refers to the accompanying drawings which illustrate, by way of illustration, specific embodiments in which the present invention may be practiced. These embodiments are described in sufficient detail to enable one skilled in the art to practice the present invention. It should be understood that the various embodiments of the present invention are different from each other but are not necessarily mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in one embodiment in another embodiment without departing from the spirit and scope of the invention. Additionally, it should be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the detailed description set forth below is not to be taken in a limiting sense, and the scope of the present invention, if properly described, is limited only by the appended claims, along with all equivalents as claimed by those claims. Like reference numbers in the drawings indicate the same or similar function throughout the various aspects.

1 is a block diagram of a modeling image recommendation model generating device 1 based on a user input keyword and artificial intelligence according to an embodiment, and FIG. 2 is a modeling image recommendation model generating device based on a user input keyword and artificial intelligence according to an embodiment. It is a flowchart for explaining the operation of (1), and FIGS. 3 to 5 are referenced to explain FIGS. 1 and 2 .

The modeling image recommendation model generation device 1 based on user input keywords and artificial intelligence may include all kinds of devices capable of generating modeling images by receiving keywords such as PCs and servers.

First, referring to FIG. 1 , an apparatus 1 for generating a modeling image recommendation model based on a user input keyword and artificial intelligence includes a control unit 10 and a storage unit 20, and the control unit 10 includes a keyword receiving unit 11 , A keyword collection unit 12, an image group reading unit 13, an image group generation unit 13-1, an image selection unit 14, a modeling unit 15, and a modeling image management unit 16 may be included. .

For reference, each order in the flowchart below is an example, and the scope of the present invention is not limited to the order.

Referring to FIGS. 1 to 5 together, the keyword reception unit 11 may receive input of a plurality of main keywords respectively indicating finished product objects classified as heterogeneous objects from the user (s1).

In the present invention, a 'main keyword' may be defined as an upper level keyword directly input by a user. That is, it may be defined as a main keyword that is a higher level concept that is a search target that a user wants to search directly.

According to the embodiment, the keyword reception unit 11 may receive input of main keywords classified into objects of different types. For example, a chair included in furniture and a bag not included in furniture are finished objects that are classified as heterogeneous objects that are not related to each other by purpose (e.g., a chair is used for sitting, a bag is used for storing things). Keywords representing each may be input as main keywords ('chair': main keyword A, 'bag': main keyword B).

The keyword collection unit 12 may collect a plurality of sub-keywords corresponding to each of a plurality of main keywords (s2).

In the present invention, a 'lower keyword' may be defined as a keyword that is a lower concept constituting a main keyword, which is a higher level concept. According to the embodiment, the lower keyword may be a keyword indicating a part object constituting a finished product object that is a target object of the main keyword. For example, sub-keywords (A1, A2 , A3), 'body' and 'handle' representing the body (Ob1) and handle (Ob2), which are part objects of the finished product object of the bag (Ob), can be collected as sub-keywords (B1 and B2), respectively.

However, this is an embodiment, and in any case where the sub-keyword is a concept included in the main keyword, it can be equally/similarly applied to the present invention.

Main keywords and sub-keywords corresponding thereto are previously databased and stored in the keyword DB 21, and the keyword collection unit 12 may search for and collect sub-keywords corresponding to the main keywords from the keyword DB 21.

According to the embodiment, the keyword collection unit 12 further determines whether or not the sub-keyword corresponding to the main keyword includes an inscription or an error and whether it does not overlap, and determines that the sub-keyword does not include an inscription or an error and does not overlap. If determined, the lower keyword may be transmitted to the image group reader 13.

The image group reading unit 13 may read a plurality of groups composed of images including a two-dimensional object corresponding to a part object indicated by each of a plurality of sub-keywords (s3).

For example, a group (C1) consisting of images (I _1-1, I _1-2?? I _1-4 ) including a 2-dimensional seat object corresponding to a seat (sub-keyword A1), a board (sub-keyword A2 ), a group (C2) consisting of images (I _2-1, I _2-2?? I _2-4 ) including a 2-dimensional climbing object corresponding to a 2-dimensional handle corresponding to a handle (sub-keyword B2) A group C5 composed of images (I _5-1, I _{5-2 ?} I _5-4) including can be read from the image group DB 22.

For reference, images corresponding to lower keywords may be images including 2D objects represented by lower keywords. That is, a real part object corresponding to a lower keyword may be expressed as a two-dimensional object and included in an image.

Each image may be mapped with a tag and a timestamp describing each image and stored in advance in the image group DB 22 by the image group generator 13-1. According to the embodiment, a tag describing an image may be generated by the image group generating unit 13-1 from metadata such as a title appearing in a web document from which an image is extracted or a word described in the text. According to an embodiment, a tag describing an image may include a tag describing a 2D object included in the image. For example, when the 2D object included in the image represents the armrest Oa3, a tag describing the size, thickness, use, and/or material of the armrest may be included. In addition, when the 2D object included in the image represents the handle Ob2, a tag describing the size, shape, use, and/or material of the handle may be included. According to embodiments, items included in tags describing 2D objects included in each image may be set to be all the same for each tag, or only some of them may be set to be the same.

According to the embodiment, the timestamp represents the creation time of the corresponding image,

If the metadata of the web document from which the corresponding image was extracted describes the production time of the real part object corresponding to the 2D object included in the image, this is set as the timestamp of the image by the image group generation unit 13-1. can be determined For example, when the 2D object included in the image represents an armrest, the actual production time of the armrest may be determined as a timestamp of the corresponding image. In addition, when the 2D object included in the image represents the handle, the actual production time of the handle may be determined by the timestamp of the corresponding image. According to an embodiment, the image group generator 13-1 may record a timestamp in a form in which all of the year, month, day, and time are reflected, so that the season in which the corresponding real part object was manufactured can be determined.

The image selector 14 may select one of the images from each of a plurality of groups (s4).

For example, an I _1-1 image from the C1 group, an I _2-1 image from the C2 group, and an I _5-1 image from the C5 group may be selected, respectively.

According to the embodiment, the image selection unit 14 selects any one image from each of the plurality of groups to indicate that each timestamp is similar to each other within a predetermined range and the season of use of each of the finished product objects is similar. You can choose.

According to an embodiment, the image selector 14 may perform the selection after arranging timestamps of images included in each of a plurality of groups in time order.

Specifically, the image selector 14 may select images having similar timestamps within a predetermined range from each group, which is to produce a real part object corresponding to the 2D object included in each image. This is to select each image indicating that the time is similar to each other within a predetermined range, and thus, to indicate that the use season of each finished product object corresponding to each real part object is similar to each other within a predetermined range Images can be combined.

For example, the image selector 14 may select an armrest image with a timestamp of January 1, 2022 and a handle image with a timestamp of January 2, 2022, which indicates an armrest whose production time is winter. This is to select an image representing an image and a handle, and as a result, an armrest image and a handle image for indicating that the season of using a chair and a bag is winter can be combined.

That is, according to the present invention, images of parts corresponding to finished products used in the same season can be combined by the image selection unit 14, so that images of unnecessary or awkward combinations can be excluded and provided to the user. This allows an effective combination process to proceed.

According to the embodiment, the image selector 14 may select one image indicating that the size of each part of each part object of each tag is similar to each other within a predetermined range from each of a plurality of groups (s4). .

Prior to going into detail, referring to FIGS. 4 and 5, a chair Oa, which is a real finished product object, includes a seat Oa1, a back plate Oa2, and an armrest Oa3, and a bag Ob has a body ( Ob1) and a handle Ob2. Also, the armrest Oa3 may include an armrest portion Oa31 and a frame portion Oa32, and the handle Ob2 may include a handle portion Ob21 and a connection portion Ob22.

Here, the image selection unit 14 may select each image indicating that the size of a portion of the armrest Oa3 and the size of a portion of the handle Ob2 are similar to each other within a predetermined range from each group. Specifically, each image indicating that the size of the armrest part Oa31 constituting the armrest Oa3 and the size of the handle part Ob21 constituting the handle Ob2 are similar to each other within a predetermined range can be selected from each group. there is. More specifically, each image showing that the width W1 of one cross section of the armrest part Oa31 and the width W2 of one cross section of the handle part Ob21 are similar to each other within a predetermined range can be selected from each group. .

That is, as a result, on the final modeling image, each image is combined so that the bag is coupled to the chair, and in particular, the handle of the bag can be displayed in a state that is suitably coupled to the armrest of the chair.

This is achieved by comparing information on the width (W1) of one end of the armrest (Oa31) and information on the width (W2) of one end of the handle (Ob21) on the tag describing each image by the image selector (14). It may be implemented by selecting each image showing similarities within this predetermined range from each group.

The modeling unit 15 may create 3D objects from the 2D objects included in each of the selected images, and generate a plurality of modeling images in which the 3D objects are combined (s5). ).

Specifically, the modeling unit 15 may generate a 3D object based on coordinates of each of the 2D objects on the image. For example, a 3D object may be created based on the positions of the points constituting each 2D object and the distance between the points.

According to an embodiment, the modeling unit 15 may generate a 3D object based on attribute data of each 2D object. Attribute data is characteristic data representing the nature and content of each 2D object, and the modeling unit 15 may create a 3D object based on the shape, texture, and color of each 2D object. As a result, it is possible to create a three-dimensional object closer to reality.

According to an embodiment, the modeling image in which each of the 3D objects is combined may represent an integrated object having a new use in a form in which the finished product objects having different uses are combined.

Specifically, keywords representing each finished product object (eg chair, bag) classified as heterogeneous objects are input, images corresponding to part objects (eg armrests, handles) are read from the corresponding sub-keywords, and selected among them are selected. The images are combined or combined so that a modeling image representing a new type of integrated object in which each of the finished product objects is combined is generated, so that when a keyword is initially input, a new association is given to heterogeneous objects that have no association by purpose ( Alternatively, a modeling image corresponding to the integrated object is provided. That is, as a result, a single modeling image corresponding to an integrated object with combined uses (e.g., a chair with a bag, a combination of uses: an object for sitting on people and a use for storing things) is provided to the user. You can do it. In other words, a new integrated object having a new purpose can be created based on a user's keyword input and image modeling.

The modeling image management unit 16 may generate an artificial intelligence-based modeling image recommendation model 24 based on the plurality of generated modeling images (s6).

Specifically, the modeling image management unit 16 provides a user with the plurality of modeling images, receives a selection input from a user through a user interface (not shown) to select some of the plurality of modeling images, and selects the selected modeling image. Some modeling images may be stored in the modeling image DB 23 together with a plurality of primary keywords initially input.

According to the present invention, an image that meets a predetermined criterion (for example, an image indicating similar use seasons, an image indicating similar size of a part, etc.) is first filtered by the device, and then filtered again by the user By storing the result of performing the difference filtering, according to the present invention, it is possible to provide a more suitable image to the user.

According to an embodiment, the modeling image management unit 16 may train a neural network model by using selected partial modeling images pre-stored in the modeling image DB 23 and a plurality of main keywords.
According to an embodiment, the modeling image management unit 16 may train a neural network model using a plurality of main keywords as training data.

delete

According to an embodiment, the neural network model (NM) may be a convolution neural network (CNN), but the present invention is the same in cases such as a deep neural network (DNN), a recurrent neural network (RNN), and a fully convolution network (FCN). /Similarly applicable.

The modeling image management unit 16 may generate an artificial intelligence-based modeling image recommendation model 24 by learning the neural network model and store it in the storage unit 20, and then receive the same main keyword from the same user or another user. In this case, the modeling image recommendation model 24 may be used to output the same partial modeling image from the same main keyword, thereby providing it to the user.

In this way, only some of the selected modeling images are selected and stored in the modeling image DB 23, and when the same main keyword is input from the user later, the user only uses some of the corresponding modeling images by using the artificial intelligence-based modeling image recommendation model 24. By providing it to the user, it is possible to provide an image suitable for the user more quickly.

The above-described embodiments may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc. alone or in combination.

Program instructions recorded on the computer-readable recording medium may be specially designed and configured for the present invention, or may be known and usable to those skilled in the art of computer software.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROMs and DVDs, and magneto-optical media such as floptical disks. optical media), and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter or the like as well as machine language codes such as those produced by a compiler. The hardware device may be configured to act as one or more software modules for executing processes according to the present invention and vice versa.

Aspects herein may take the form of entirely hardware, entirely software (including firmware, resident software, microcode, etc.) or a computer program product embodied entirely in one or more computer readable media having computer readable program code embodied thereon. .

Features, structures, effects, etc. described in the embodiments above are included in one embodiment of the present invention, and are not necessarily limited to only one embodiment. Furthermore, the features, structures, and effects illustrated in each embodiment can be combined or modified with respect to other embodiments by those skilled in the art in the field to which the embodiments belong. Therefore, contents related to these combinations and variations should be construed as being included in the scope of the present invention.

In addition, although the above has been described with a focus on the embodiments, these are only examples and do not limit the present invention, and those skilled in the art to which the present invention belongs can exemplify the above to the extent that does not deviate from the essential characteristics of the present embodiment. It will be seen that various variations and applications that have not been made are possible. For example, each component specifically shown in the embodiment can be modified and implemented. And differences related to these modifications and applications should be construed as being included in the scope of the present invention as defined in the appended claims.

Claims (7)

a keyword receiving unit that receives a plurality of main keywords representing finished product objects classified as heterogeneous objects from a user;
a keyword collection unit for collecting the plurality of sub-keywords corresponding to each of the plurality of main keywords by referring to a database in which a plurality of sub-keywords corresponding to each of the plurality of main keywords are stored in advance;
an image group reading unit for reading a plurality of groups composed of images including a two-dimensional object corresponding to a part object indicated by each of the plurality of sub-keywords;
an image selector selecting one of the images from each of the plurality of groups;
a modeling unit generating a plurality of modeling images based on each of the selected images; and
A modeling image management unit generating an artificial intelligence-based modeling image recommendation model based on the plurality of generated modeling images;
The image selection unit,
Selecting any one image from each of the plurality of groups indicating that the size of each part of each part object of the tag of each image is similar to each other within a predetermined range,
Modeling image recommendation model generation device based on user input keywords and artificial intelligence.
delete delete delete According to claim 1,
The modeling image management unit,
The plurality of modeling images are provided to the user, some of the plurality of modeling images are selected and input from the user, and a neural network model is trained using the selected modeling images and the plurality of main keywords. , Learning the neural network model to generate the artificial intelligence-based modeling image recommendation model,
Modeling image recommendation model generation device based on user input keywords and artificial intelligence.
According to claim 1,
The modeling image represents an integrated object having a new use in the form of a combination of the finished product object having a different use,
Modeling image recommendation model generation device based on user input keywords and artificial intelligence.
a keyword receiving unit that receives a plurality of main keywords representing finished product objects classified as heterogeneous objects from a user;
a keyword collection unit for collecting the plurality of sub-keywords corresponding to each of the plurality of main keywords by referring to a database in which a plurality of sub-keywords corresponding to each of the plurality of main keywords are stored in advance;
an image group reading unit for reading a plurality of groups composed of images including a two-dimensional object corresponding to a part object indicated by each of the plurality of sub-keywords;
an image selector selecting one of the images from each of the plurality of groups;
a modeling unit generating a plurality of modeling images based on each of the selected images; and
A modeling image management unit generating an artificial intelligence-based modeling image recommendation model based on the plurality of generated modeling images;
The image selection unit,
The timestamps of each of the images are similar to each other within a predetermined range to select the one image indicating that the production time of each of the part objects is similar to each other within a predetermined range from each of the plurality of groups,
Modeling image recommendation model generation device based on user input keywords and artificial intelligence.

Images