KR102674699B1 - Deep learning-based 3D modeling object visualization apparatus and method - Google Patents

Abstract

A deep learning-based 3D modeling object visualization device and method are provided. The setting unit places the 3D modeling object in the 3D background space and sets the light properties of the 3D background space, and the deep learning learning unit applies the metadata of the 3D modeling object and the light properties of the 3D background space to a deep learning algorithm to determine the light properties. It creates an inference model that automatically adjusts polygons and textures accordingly, the rendering unit renders 3D modeling objects with polygons and textures inferred by the generated inference model, and the visualization unit creates a 3D background space where the rendered 3D modeling objects are placed. can be visualized.

Description

Deep learning-based 3D modeling object visualization apparatus and method}

The present invention relates to a deep learning-based 3D modeling object visualization device and method. More specifically, a deep learning-based 3D modeling object that can be visualized by automatically inferring and applying polygons and textures to be applied to 3D objects according to light. It relates to a visualization device and method.

Recently, as various application fields using virtual reality or augmented reality have been in the spotlight, fast and accurate 3D model creation is required.

Typically, to create a 3D virtual object, workers analyze various properties of the physical object, such as material, color, and size, and model the 3D virtual object individually and manually. At this time, the operator must create additional polygons and textures for modeling, taking into account the material, color, and size of the physical object. Therefore, producing 3D virtual objects using existing methods requires a lot of time and money.

In addition, a polygon is a polyhedron composed of a set of points (vertex), lines (edge), and faces (polygon) that make up the shape, and is structured very intuitively. However, because the expression of curves is insufficient, aliasing occurs, and to express curves with polygons, the number of polygons must be increased, which requires high hardware (graphics card) performance. In particular, when creating a mesh, which is a collection of polygons, modeling work must be done considering optimization according to hardware constraints and design concepts.

Domestic Public Patent No. 10-2015-0124208

In order to solve the above-mentioned problems, the technical problem to be achieved by the present invention is to present a deep learning-based 3D modeling object visualization device and method that can automatically infer and apply polygons and textures to be applied to 3D objects according to light and visualize them. There is something to do.

The problems to be solved by the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below.

As a means to solve the above-described technical problem, according to an embodiment of the present invention, a deep learning-based 3D modeling object visualization device places a 3D modeling object in a 3D background space and sets light properties of the 3D background space. Setting section; A deep learning learning unit that generates an inference model that automatically adjusts polygons and textures according to light properties by applying metadata of the 3D modeling object and light properties of the 3D background space to a deep learning algorithm; A rendering unit that renders a 3D modeling object to which polygons and textures inferred by the generated inference model are applied; and a visualization unit that visualizes a 3D background space where the rendered 3D modeling object is placed.

The deep learning learning unit includes a classification unit that classifies the 3D modeling object into a plurality of detailed objects; an inference model generator for generating an inference model for each detailed object by learning background data including metadata of each detailed object classified by the classification unit and light properties of the 3D background space based on the deep learning algorithm; and an application unit that infers and applies polygon and texture information to be applied to each detailed object according to light properties, using an inference model generated for each detailed object.

The classification unit determines whether there are errors in the plurality of detailed objects and classifies them into normal objects and defective objects, and the inference model creation unit may generate an inference model when the 3D modeling object does not include a defective object.

The application unit may infer and apply the tension value, number, and texture of the polygon according to the currently set light properties, using the inference model generated for each detailed object.

The setting unit may set light properties including at least one of light intensity, thickness, and color temperature to be applied to the 3D background space.

The metadata of the 3D modeling object includes texture, size, and detailed attribute information, and the detailed attribute information is applied to each detailed object of the 3D modeling object according to the intensity of light, thickness of light, direction of light, and color temperature of light. This is data that maps the tension, number, and texture information of the polygons to be processed.

Meanwhile, according to another embodiment of the present invention, a deep learning-based 3D modeling object visualization method includes the steps of: (A) an electronic device, placing a 3D modeling object in a 3D background space and setting light properties of the 3D background space; ; (B) generating, by the electronic device, an inference model that automatically adjusts polygons and textures according to light properties by applying metadata of the 3D modeling object and light properties of the 3D background space to a deep learning algorithm; (C) rendering, by the electronic device, a 3D modeling object to which polygons and textures inferred by the generated inference model are applied; and (D) visualizing, by the electronic device, a 3D background space where the rendered 3D modeling object is placed.

Step (B) includes: (B1) classifying the 3D modeling object into a plurality of detailed objects; (B2) generating an inference model for each detailed object by learning background data including metadata of each detailed object classified in step (B1) and light properties of the 3D background space based on the deep learning algorithm; And (B3) using the inference model generated for each detailed object in step (B2), inferring and applying polygon and texture information to be applied to each detailed object according to light properties.

In the step (B1), the plurality of detailed objects are judged to have errors and are classified into normal objects and defective objects, and in the step (B2), when the 3D modeling object does not include a defective object, an inference model is generated. can do.

In step (A), light properties including at least one of light intensity, thickness, and color temperature to be applied to the 3D background space may be set.

According to the present invention, by automatically inferring the polygon and texture information of a 3D modeling object according to the light properties, the user avoids the inconvenience of manually selecting the tension or number of polygons and the texture of the 3D modeling object every time the light property changes. The feeling can be relieved.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 is a block diagram showing a deep learning-based 3D modeling object visualization device 100 according to an embodiment of the present invention.
Figure 2 is a conceptual diagram showing a series of operations for rendering 3D modeling objects by learning deep learning.
Figure 3 is a block diagram functionally showing the processor 150 shown in Figure 1;
Figure 4 is an example diagram for explaining the mesh of a 3D modeling object, and
Figure 5 is a flowchart schematically showing a deep learning-based 3D modeling object visualization method according to an embodiment of the present invention.

The above objects, other objects, features and advantages of the present invention will be easily understood through the following preferred embodiments related to the attached drawings. However, the present invention is not limited to the embodiments described herein and may be embodied in other forms. Rather, the embodiments introduced herein are provided so that the disclosed content will be thorough and complete and so that the spirit of the present invention can be sufficiently conveyed to those skilled in the art.

In some cases, it is mentioned in advance that when describing the invention, parts that are commonly known but are not significantly related to the invention are not described in order to prevent confusion without any reason in explaining the invention.

In this specification, when it is mentioned that a first component is operating or executing on a second component (ON), the first component is operating or executing in an environment in which the second component is operating or executing, or is operating or executing on the second component. It should be understood as operating or executing through direct or indirect interaction with.

When a component, device, or system is said to contain a component consisting of a program or software, even if explicitly stated, that component, device, or system refers to the hardware (hardware) necessary for the program or software to run or operate. For example, memory, CPU, etc.) or other programs or software (for example, drivers necessary to run an operating system or hardware, etc.).

Additionally, unless otherwise specified, it should be understood that the component may be implemented in any form of software, hardware, or both software and hardware.

Additionally, the terms used in this specification are for describing embodiments and are not intended to limit the present invention. As used herein, singular forms also include plural forms, unless specifically stated otherwise in the context. As used in the specification, 'comprises' and/or 'comprising' does not exclude the presence or addition of one or more other elements.

Additionally, in this specification, terms such as 'unit' and 'device' may be intended to refer to the functional and structural combination of hardware and software driven by or for driving the hardware. For example, the hardware here may be a data processing device including a CPU or other processor. Additionally, software driven by hardware may refer to a running process, object, executable, thread of execution, program, etc.

In addition, the above terms may mean a logical unit of hardware resources for executing a predetermined code and the predetermined code, and do not necessarily mean a physically connected code or one type of hardware. It can be easily inferred by an average expert in the technical field.

Hereinafter, specific technical details to be implemented in the present invention will be described in detail with reference to the accompanying drawings.

Each configuration shown in FIGS. 1 and 2 indicates that it may be functionally and logically separated, and does not necessarily mean that each configuration is divided into a separate physical device or written in a separate code. The average expert in the technical field will be able to infer easily.

Figure 1 is a block diagram showing a deep learning-based 3D modeling object visualization device 100 according to an embodiment of the present invention, and Figure 2 is a conceptual diagram showing a series of operations for rendering a 3D modeling object by deep learning learning.

Referring to FIG. 1, the deep learning-based 3D modeling object visualization device 100 according to an embodiment of the present invention includes a user interface unit 110, a display unit 120, a storage unit 130, a memory 140, and a processor ( 150) may be included.

The user interface unit 110 provides an interfacing path between the user and the deep learning-based 3D modeling object visualization device 100, and can receive commands from the user and transmit them to the processor 150. For example, the user interface unit 110 may select a 3D modeling object to be visualized in 3D, select a 3D background space in which to place the 3D modeling object, and transmit the selection to the processor 150 according to a user command. Examples of the user interface unit 110 include a mouse, keyboard, and touch panel.

The display unit 120 may display a processing result of the processor 150 according to a user command of the user interface unit 110. As an example, the display unit 120 may display a 3D background space in which a 3D modeling object is placed, a 3D modeling object to be placed in the 3D background space, and a 3D virtualization object placed in the 3D background space.

The storage unit 130 may store a plurality of 3D background spaces, that is, background images, in which 3D modeling objects are to be placed. The 3D background space may be, for example, a park, a city with high-rise buildings and roads, or a riverside, but this is not limited to examples.

Additionally, the storage unit 130 may store objects to be deep-learned in the deep learning-based 3D modeling object visualization device 100, that is, virtual 3D modeling objects and metadata of the 3D modeling objects for each plurality of background data.

More specifically, metadata may be stored in the storage unit 130 for each of the multiple detailed objects that make up the 3D modeling object. For example, the 3D modeling object shown in FIG. 2 is a house and includes a number of detailed objects such as a roof, windows, entrance, walls, and pillars. In this case, metadata is stored in the storage unit 130 for each detailed object such as a roof, window, entrance, wall, and pillar, and may be mapped and stored for each plurality of background data.

Background data includes the intensity and thickness of light to be applied (or set) to the 3D background space as basic information, and may further include directions of light and color temperatures of light as additional information. The number of background data can be adjusted according to the combination of light intensity, thickness, direction, or color temperature.

The metadata of a 3D modeling object is data that maps the tension, number, and texture information of polygons to be applied to each detailed object of the 3D modeling object according to the background data, and includes attribute information necessary to visualize the 3D modeling object.

The intensity of light is the intensity of light set in the 3D background space where the 3D modeling object is to be placed, and can be defined as levels 1 to 12, for example. The larger the number, the brighter the light can be.

The thickness of light is the thickness of light set in the 3D background space and can be defined, for example, in levels 1 to 10. The larger the number, the stronger the light can be.

The direction of light is the incident direction or incident angle of light and can be set in units specified by the user, such as 10° or 30° units within 0 to 360°.

The color temperature of light is information representing the color of light and can be defined as bulb color, daylight color, daylight white, etc. The color temperature of light can be set differently depending on whether the 3D background space is indoors or outdoors.

Polygon tension is the size of the polygon to be applied to each detailed object that makes up a 3D modeling object. Even for the same detailed object, the tension of a polygon can be set to different values depending on the intensity, thickness, direction, and color temperature of the light. For example, if at least one of the light intensity and thickness set in the 3D background space is weak and dark, and the light is incident from a 20° direction, the bottom of the roof does not need to be expressed in detail, so a polygon of the same size as the maximum detail object is used. Can be set to use.

The number of polygons is the number of polygons to be applied to each sub-object considering the tension of the polygon.

The texture may include at least one of color, brightness, and texture information to be applied to the polygon of the sub-object. The color of the texture can be set in the RGB range of 0 to 255, and the brightness can be set in the range of 0 (black) to 10 (white). Texture may be material information of a sub-object. Materials vary, including wood, metal, plastic, stone, and glass.

[Table 1] is an example of metadata set for each detailed object when the light intensity and light thickness among the background data are 1 and the light direction is 45°.

background data metadata light intensity thickness of light light direction Detailed object Material (texture) Size (cm) polygon tension number of polygons texture One One 45° window 1 glass 30X50 Same as window 1 One Color: (85, 20, 0)
Brightness: 7 entrance metal 150X200 1/4 of the vestibule 4 Color: (202, 0, 0)
Brightness: 8 wall 1 wood 300X400 1/10 of wall 1 10 Color: (20, 94, 25) Brightness: 5

[Table 2] is an example of metadata set for each detailed object when the light intensity is 2, the light thickness is 1, and the light direction is 90° among the background data.

background data metadata light intensity thickness of light light direction Detailed object Material (texture) size
(cm) polygon
tension number of polygons texture 2 One 90° window 1 glass 30X50 1/2 of window 1 2 Color: (100, 20, 0)
Brightness: 7 entrance metal 150X200 Same as the entrance One Color: (255, 0, 0) Brightness: 8 wall 1 wood 300X400 1/2 of wall 1 2 Color: (20, 204, 255)
Brightness: 5

Referring to [Table 1] and [Table 2], you can see that even for the same detailed object, polygons and textures are set differently depending on background data such as light intensity, thickness, and direction. Accordingly, the metadata of a detailed object may be changed depending on the number of cases generated by a combination of background data (light intensity, thickness, direction, color temperature, etc.).

Memory 140 may include volatile memory and/or non-volatile memory. The memory 140 includes, for example, commands or data related to the components 110 to 150 in order to implement and/or provide operations, functions, etc. provided by the deep learning-based 3D modeling object visualization device 100, One or more programs and/or software, operating systems, etc. may be stored.

The program stored in the memory 140 is a machine that deep learns the default values of the background data and the metadata of the 3D modeling object to create an inference model that can automatically infer the polygon and texture of the 3D modeling object according to the properties of light. May include a learning algorithm.

Additionally, the memory 140 may store a rendering program that renders and visualizes 3D modeling objects using polygons and textures generated by the inference model.

The processor 150 executes one or more programs stored in the memory 140 to control the overall operation of the deep learning-based 3D modeling object visualization device 100.

For example, the processor 150 executes a machine learning algorithm stored in the memory 140 to generate an inference model that automatically infers the polygons and textures of a 3D modeling object according to the properties of light, and executes a rendering program to create a 3D model. After rendering modeling objects, they can be processed to be displayed by placing them in the 3D background space.

FIG. 3 is a functional block diagram of the processor 150 shown in FIG. 1.

Referring to FIG. 3, the processor 150 may include a setting unit 152, a deep learning learning unit 154, a rendering unit 156, and a visualization unit 158.

When the user operates the user interface unit 110 and selects a 3D modeling object and a 3D background space, the setting unit 152 places the 3D modeling object at a location designated by the user among the selected 3D background spaces, and places the 3D modeling object in the location selected by the user. You can set the light properties to be applied to the 3D background space.

When a user requests setting the light properties of a 3D background space, the setting unit 152 creates a setting menu (not shown) showing the settable light intensity and thickness range, light direction, and color temperature and displays it on the display unit 120. can do. The user essentially selects the intensity and thickness of light to be applied to the 3D background space through the settings menu (not shown), and can omit or additionally select the light direction and color temperature. This is because polygons and textures among metadata are most affected by the intensity and thickness of light among the properties of light. The setting unit 152 may store the light properties set in the 3D background space in the memory 140 as default background data.

The deep learning learning unit 154 applies and learns the metadata of the selected 3D modeling object and the light properties (i.e., background data) of the set 3D background space to a deep learning algorithm to automatically infer polygons and textures according to the light properties. A capable inference model can be created.

The deep learning training unit 154 may include a classification unit 154a, an inference model creation unit 154b, and an application unit 154c.

The classification unit 154a may classify 3D modeling objects into multiple detailed objects. At this time, the classification unit 154a may determine whether there are errors in a number of detailed objects and classify them into normal objects and defective objects.

In detail, the classification unit 154a may extract the coordinates of each detailed object constituting the 3D modeling object, divide the coordinates into pixel-unit images, and store them in the memory 140. The classification unit 154a performs a 3D virtual object geometric analysis technique, that is, a process of extracting classification, components (Parsing/Segmentation), and correlation information between sub-objects of the input 3D modeling object. Modeling objects can be classified and categorized by subdividing them into detailed objects. If the 3D modeling object is a house, detailed objects may include roofs, windows, doors, walls, floors, pillars, stairs, etc.

In addition, in this classification operation, the classification unit 154a separates detailed objects into normal components without defects (normal objects) and components with defects (defective objects), and then sorts them according to the type of defect or edges ( Detailed objects can be assigned to different classes, such as assigning detailed objects (or 3D modeling objects) without edges to different categories.

For example, in the task of modeling a house as a 3D modeling object, the classification unit 154a performs the task of checking whether a problem has occurred in the 3D house or whether a part of the 3D modeled house is broken or missing. You can proceed with Categorization.

The operation of extracting 3D modeling objects with defects described above is to minimize errors in which incorrectly created virtualization models (i.e., 3D modeling objects) may cause metadata values to be misinterpreted during the machine learning algorithm learning/analysis process. am. For example, if one edge of a window is missing, the color of the polygon is connected to another detailed object connected to the window, resulting in light spreading or the detailed object being unable to be clearly distinguished.

If there is no defective object, the classification unit 154a may assign the mesh of the 3D modeling object to one class as shown in FIG. 4. Through this, the classification unit 154a can perform classification and identification of multiple 3D detailed objects consisting of vertices, polygons, and edges of 3D modeling objects.

The inference model creation unit 154b may generate an inference model when the 3D modeling object in the classification unit 154a does not include a defect object.

The inference model creation unit 154b generates the metadata of each detailed object classified in the classification unit 154a, the light properties of the 3D background space (i.e., default background data) set in the setting unit 152, and the light properties in the 3D background space. By learning background objects (for example, trees in a park, street lights, buildings, etc.) based on a deep learning algorithm, an inference model can be created for each detailed object. Background objects may be omitted.

The inference model is an AI model that can automatically infer the tension, number, and texture of polygons to be applied to the 3D modeling object for each detailed object according to the changed light properties when the user changes the light properties. This is because even with the same light properties, the polygons and textures applied to the detailed objects may be different depending on the material of the detailed object, the location where the detailed object is located, etc.

Additionally, the inference model generator 154b may repeatedly perform learning to generate an inference model using the position (X, Y, Z) where the user places the 3D modeling object in the 3D background space as label information. In detail, once the inference model is created, the user may request to correct the metadata of the 3D modeling object or change the position where the 3D modeling object is placed in the 3D background space and then create or update the inference model again.

For example, if the inference result of the 3D virtualized object visualized on the display unit 120 is different from the result desired by the user, the user may correct the metadata of the 3D modeling object and then rearrange the 3D modeling object in the 3D background space. there is. The classification unit 154a classifies the relocated 3D modeling object into detailed objects and determines whether there is an error, and the inference model creation unit 154b labels the location of the relocated 3D modeling object and then regenerates or updates the inference model. You can.

The application unit 154c may use an inference model generated for each detailed object to infer polygon and texture information to be applied to each detailed object according to light properties and apply them to the detailed object. When the inference model is first created, the application unit 154c can infer the tension, number, and texture of polygons corresponding to the light properties set as default in the setting unit 152 for each detailed object.

In addition, after the inference model is created, if the user changes the light properties to be applied to the 3D background space in the settings menu (not shown), the setting unit 152 resets the light properties to the changed values, and the application unit 154c Using the inference model created for each detailed object, the tension value, number, and texture of polygons can be inferred and applied according to the currently set light properties.

The rendering unit 156 may generate a 3D virtualized object by rendering each detailed object to which the polygon and texture inferred by the inference model in the application unit 154c are applied, that is, a 3D modeling object.

The visualization unit 158 can place the generated 3D virtual object in the 3D background space, visualize it, and display it on the display unit 120 as shown in FIG. 2 .

According to the above-described embodiment of the present invention, the characteristics (metadata) of 3D modeling objects are extracted and classified using a deep learning algorithm, an inference model is created, and the face of the polygon is adjusted according to the intensity and thickness of light through repeated learning. It automatically generates the tension value of the mesh that can be made soft, automatically adjusts the number of polygons, and automatically maps the color and texture to a texture that matches the properties of the 3D modeling object to render it as an optimal 3D virtualization object.

In addition, according to the above-described embodiment of the present invention, if the 3D background space changes even if the user selects the 3D modeling object shown in FIG. 2, the deep learning-based 3D modeling object visualization device 100 displays in the changed 3D background space. You can re-create an inference model to infer the polygons and textures of 3D modeling objects based on light properties.

Figure 5 is a flowchart schematically showing a deep learning-based 3D modeling object visualization method according to an embodiment of the present invention.

The electronic device that performs the deep learning-based 3D modeling object visualization method shown in FIG. 5 may be the deep learning-based 3D modeling object visualization device 100 described with reference to FIGS. 1 to 4 . Therefore, detailed description of the deep learning-based 3D modeling object visualization method is omitted.

Referring to FIG. 5 , the electronic device 100 may place a 3D modeling object selected through the user interface unit 110 in the 3D background space and display it on the display unit 120 (S510).

The electronic device 100 may set the light properties of the 3D background space selected in the settings menu as default background data through the user interface unit 110 (S520). Light properties may include the intensity and thickness of light to be applied to the 3D background space, light direction, and color (or color temperature).

The electronic device 100 can read the metadata of the 3D modeling object selected in step S510 from the storage unit 130 and store it in the memory 140 or a separate storage medium (not shown) along with the light properties of the 3D background space. There is (S530).

The electronic device 100 may classify the 3D modeling object selected in step S510 into a plurality of detailed objects (S540).

The electronic device 100 learns background data including the metadata of each detailed object classified in step S540 and the light properties of the 3D background space set in step S520 based on a deep learning algorithm to create an inference model for each detailed object. (S550, S560).

The electronic device 100 uses the inference model generated for each detailed object in step S560 to infer the tension value and number of polygons to be applied to each detailed object and texture information according to light properties, and then uses the inferred polygon and texture. It can be applied to the corresponding detailed object (S570).

The electronic device 100 may generate a 3D virtualized object by rendering the 3D modeling object to which the polygons and textures inferred in step S570 are applied (S580).

The electronic device 100 may arrange and visualize the 3D modeling object rendered in step S580 in the 3D background space and display it on the display unit 120 (S590).

After the 3D virtualized object is displayed on the display unit 120, when the user requests to correct some of the metadata of the 3D modeling object or change the location where the 3D modeling object is placed in the 3D background space, the electronic device 100 performs the above-described You can repeat the operation to re-create the inference model or update a previously created inference model.

In the above, even though all the components constituting the embodiment of the present invention have been described as being combined or operated in combination, the present invention is not necessarily limited to this embodiment. That is, as long as it is within the scope of the purpose of the present invention, all of the components may be operated by selectively combining one or more of them. In addition, although all of the components may be implemented as a single independent hardware, a program module in which some or all of the components are selectively combined to perform some or all of the combined functions in one or more pieces of hardware. It may also be implemented as a computer program with . The codes and code segments that make up the computer program can be easily deduced by a person skilled in the art of the present invention. Such a computer program can be stored in a computer-readable storage medium and read and executed by a computer, thereby implementing embodiments of the present invention.

Meanwhile, although the preferred embodiments have been described and illustrated to illustrate the technical idea of the present invention, the present invention is not limited to the configuration and operation as shown and described, and does not deviate from the scope of the technical idea. Without limitation, those skilled in the art will understand that many changes and modifications can be made to the present invention. Accordingly, all such appropriate changes, modifications and equivalents should be considered to fall within the scope of the present invention. Therefore, the true scope of technical protection of the present invention should be determined by the technical spirit of the attached registration claims.

100: Deep learning-based 3D modeling object visualization device
110: user interface unit 120: display unit
130: storage unit 140: memory
150: Processor 152: Setting unit
154: Deep learning learning unit 156: Rendering unit
158: Visualization department

Claims (11)

a setting unit that places a 3D modeling object in a 3D background space and sets light properties of the 3D background space;
A deep learning learning unit that generates an inference model that automatically adjusts polygons and textures according to light properties by applying metadata of the 3D modeling object and light properties of the 3D background space to a deep learning algorithm;
A rendering unit that renders a 3D modeling object to which polygons and textures inferred by the generated inference model are applied; and
A visualization unit that visualizes the 3D background space in which the rendered 3D modeling object is placed,
The deep learning department
Extract the coordinates of each detailed object that makes up the 3D modeling object, divide it into a pixel-unit image and store it in memory, classify and categorize the 3D modeling object into detailed objects by extracting correlation information between classification, components, and sub-objects. a classification unit that does;
an inference model generator for generating an inference model for each detailed object by learning background data including metadata of each detailed object classified by the classification unit and light properties of the 3D background space based on the deep learning algorithm; and
Using the inference model created for each detailed object, the polygon tension value and number to be applied to each detailed object and the color, brightness, and texture information to be applied to the polygon according to the light intensity, thickness, direction, and color temperature for the same detailed object. an application unit that infers and applies texture information including one;
A deep learning-based 3D modeling object visualization device comprising a.
delete According to paragraph 1,
The classification department,
Determine whether there are errors in the plurality of detailed objects and classify them into normal objects and defective objects,
The inference model generator,
A deep learning-based 3D modeling object visualization device that generates an inference model when the 3D modeling object does not include a defective object.
delete According to paragraph 1,
The settings section,
A deep learning-based 3D modeling object visualization device, characterized in that setting light properties including at least one of light intensity, thickness, and color temperature to be applied to the 3D background space.
According to paragraph 1,
The metadata of the 3D modeling object includes texture, size, and detailed attribute information,
The detailed attribute information is,
Deep learning-based 3D modeling object visualization, characterized in that it is data that maps the tension, number, and texture information of polygons to be applied to each detailed object of the 3D modeling object according to the intensity of light, thickness of light, direction of light, and color temperature of light. Device.
(A) placing, by an electronic device, a 3D modeling object in a 3D background space and setting light properties including at least one of light intensity, thickness, and color temperature to be applied to the 3D background space;
(B1) Extract the coordinates of each detailed object from the 3D modeling object, divide it into pixel-unit images and store them in memory, extract the correlation information between classification, components, and sub-objects of the 3D modeling object, and classify it into detailed objects. The step of classifying into categories;
(B2) The tension and number of polygons to be applied to each detailed object of the 3D modeling object, and the color to be applied to the polygon, according to the texture, size, light intensity, thickness, direction, and color temperature of each detailed object classified in step (B1) above. , texture containing one of brightness and texture information, metadata containing data mapping at least one of brightness information, and background data containing light properties of the 3D background space are learned based on a deep learning algorithm to create each detailed object. generating a separate inference model; and
(B3) using the inference model generated for each detailed object in step (B2), inferring and applying polygon and texture information to be applied to each detailed object according to light properties;
(C) rendering, by the electronic device, a 3D modeling object to which polygons and textures inferred by the generated inference model are applied; and
(D) visualizing, by the electronic device, a 3D background space where the rendered 3D modeling object is placed;
Deep learning-based 3D modeling object visualization method including.
delete In clause 7,
In step (B1),
Determine whether there are errors in the plurality of detailed objects and classify them into normal objects and defective objects,
In step (B2),
A deep learning-based 3D modeling object visualization method, characterized in that generating an inference model when the 3D modeling object does not include a defective object.
delete delete

Images