3D scene generation method based on large language model, 3D scene generation device, electronic device, and storage medium

Abstract

To provide a 3D scene generation method based on a large language model, which is a technical field such as a 3D modeling and a large language model, a 3D scene generation device, an electronic device, and a storage medium.SOLUTION: A 3D scene generation method includes the steps of: processing description information of a target 3D scene to acquire label information in the description information; and generating query operation presentation information of the large language model (LLM) on the basis of the label information and acquiring a target asset set matched with the label information on the basis of the query operation presentation information using LLM. The target asset set includes a target asset in the target 3D scene, target material information of the target asset and target scene attribute information of the target asset. The method also includes a step of generating a target 3D scene on the basis of the target asset set.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to the field of 3D modeling, large-scale models, etc., and in particular to a method, device, apparatus, and storage medium for generating 3D scenes based on large-scale models. [Background technology]

[0002] With the development of artificial intelligence technology, three-dimensional (3D) scene creation has been widely applied in the game industry, animation industry, digital human industry, etc.

[0003] In the related art, three-dimensional scenes are typically artificially created by a designer in a rendering engine. Summary of the Invention [Problem to be solved by the invention]

[0004] The present disclosure provides a method, apparatus, device, and storage medium for generating a three-dimensional scene based on a large-scale model. [Means for solving the problem]

[0005] According to one aspect of the present disclosure, a method for generating a 3D scene based on a large-scale model is provided, comprising: processing description information of a target 3D scene to obtain label information in the description information; generating query operation presentation information of an LLM based on the label information; and using the LLM to obtain a target asset set matching the label information based on the query operation presentation information, wherein the target asset set includes target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets; and generating the target 3D scene based on the target asset set.

[0006] According to another aspect of the present disclosure, a 3D scene generation device based on a large-scale model is provided, comprising: a processing module for processing description information of a target 3D scene to obtain label information in the description information; an acquisition module for generating query operation presentation information of an LLM based on the label information and using the LLM to obtain a target asset set matching the label information based on the query operation presentation information, wherein the target asset set includes target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets; and a generation module for generating the target 3D scene based on the target asset set.

[0007] According to another aspect of the present disclosure, there is provided an electronic device comprising at least one processor and a memory communicatively coupled to the at least one processor, the memory storing instructions executable by the at least one processor, the instructions being executable by the at least one processor to cause the at least one processor to perform the method of the aspect and any possible implementation manner.

[0008] According to another aspect of the present disclosure, there is provided a non-transitory computer readable storage medium having computer instructions stored thereon, the computer instructions causing the computer to perform the method of the aspect and any possible implementations thereof.

[0009] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method of the aspect and any possible implementation manner.

[0010] According to the technical solution of the present disclosure, a three-dimensional scene can be generated simply and efficiently.

[0011] It should be understood that the contents described herein are not intended to identify key or important features of the embodiments of the present disclosure, nor are they used to limit the scope of the present disclosure. Other features of the present disclosure can be readily understood through the following specification. [Brief description of the drawings]

[0012] The drawings are for a better understanding of the present application and are not intended to limit the present application. [Figure 1] FIG. 1 is a schematic diagram according to a first embodiment of the present disclosure. [Diagram 2] FIG. 1 is a schematic diagram for realizing an application scene of an embodiment of the present disclosure. [Diagram 3] 1 is a schematic diagram of the overall architecture of a large-scale model-based 3D scene generation system provided by an embodiment of the present disclosure. [Figure 4] FIG. 2 is a schematic diagram according to a second embodiment of the present disclosure. [Diagram 5] FIG. 11 is a schematic diagram according to a third embodiment of the present disclosure. [Figure 6] FIG. 1 is a schematic diagram of an electronic device for implementing a large-scale model-based three-dimensional scene generation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0013] Hereinafter, exemplary embodiments of the present application will be described based on the drawings. For ease of understanding, various details of the embodiments of the present application are included, and they should be regarded as merely examples. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. Similarly, for the sake of simplicity, the following description omits the description of well-known functions and structures.

[0014] In the related technologies, 3D scenes are artificially created, so there are problems in terms of efficiency, accuracy, etc.

[0015] To improve the accuracy and efficiency of 3D scene generation, this disclosure provides the following examples.

[0016] 1 is a schematic diagram according to a first embodiment of the present disclosure. This embodiment provides a method for generating a 3D scene based on a large-scale model, the method including the following steps: 101, process description information of a target 3D scene to obtain label information in the description information.

[0017] 102. Generate query operation presentation information of an LLM based on the label information, and use the LLM to obtain a target asset set matching the label information based on the query operation presentation information, wherein the target asset set includes target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets.

[0018] 103, generating the target 3D scene based on the target asset set.

[0019] The description information refers to the description information for the target 3D scene input by a user. For example, the description information is "I want to generate a scene of playing in a forest in northern China in winter. In the scene, there is a dirt road and a house made of piles, white smoke is coming out of the chimney on the roof of the house, and warm yellow lights are on in the windows of the house. The color is cool and realistic."

[0020] The label information is key information for the description information. For example, based on the above example, the label information includes: a scene, winter season, northern China, forest, for games, a dirt road, a house made of piles, a chimney on the roof of the house, white smoke coming out of the chimney, and a warm yellow light in the window of the house, with cold tones and realistic.

[0021] After obtaining the label information, a large language model (LLM) can be used to obtain a target asset set that matches the label information. The input of the LLM includes a prompt, and the LLM generates corresponding output information based on the prompt.

[0022] To obtain the target asset set, the presentation information of the LLM can be called a query operation presentation information, and the LLM queries and obtains a target asset set that matches the label information based on the query operation presentation information. Wherein, the query operation presentation information can include the above-mentioned label information, and can also include instruction information, which is used to instruct the LLM to perform a query operation, and the specific content of the instruction information can be set. After receiving the query operation presentation information, the LLM can obtain the target asset set based on the presentation information.

[0023] A 3D scene is made up of assets, which include, for example, models, material balls, maps, particles, etc., and models include, for example, models of houses, flowers, grass, etc.

[0024] An asset that matches the target 3D scene is called a target asset. For example, if the label information includes a house, a house model is obtained as the target asset.

[0025] The target scene attribute information refers to the relevant information of the target asset in the target 3D scene, including, for example, size, position, rotation angle, and so on.

[0026] Target material information refers to material-related information of the target asset, such as color, roughness, etc.

[0027] After obtaining the target asset set, a target 3D scene can be generated based on the target asset set, for example, by generating a target asset corresponding to the size based on the size information in the scene attribute information, and placing the generated target asset at a specified position in the scene attribute information.

[0028] In the related technology, when only the target asset is obtained, the scene attribute information of the target asset is undetermined, so for example, the position of the model is randomly determined, which requires the user to make further adjustments to the target asset, which increases complexity.

[0029] In this embodiment, the target asset set not only includes the target asset, but also includes scene attribute information and material information of the target asset, which corresponds to obtaining the target asset and its material, as well as scene attribute information such as position, size, etc. at one time, and can generate a 3D scene more simply and efficiently than the method of obtaining only the target asset. In addition, by generating the query operation presentation information of the LLM, the target asset set can be obtained using the LLM, which improves the processing efficiency and accuracy.

[0030] In order to better understand the embodiments of the present disclosure, application scenarios to which the embodiments of the present disclosure can be applied will be described.

[0031] 2 is a schematic diagram of an application scenario provided by an embodiment of the present disclosure, which includes a user terminal 201 and a server 202, where the user terminal 201 includes a personal computer (PC), a mobile device (e.g., a mobile phone), a tablet, a notebook computer, a smart wearable device, etc., the server 202 can be a local server or a cloud server, etc., and the server can be a single server or a server cluster.

[0032] As shown in FIG. 3, the 3D scene generation system based on a large model includes a rendering engine 301, a plug-in 302, a large model (Large Language Model, LLM) 303, and an asset repository. The rendering engine 301 is configured in a user terminal, and the rendering engine is, for example, Unity software, virtual engine (Unreal Engine, UE) software, etc., and the Unity software and the UE software (specifically may be UE4 or UE5) are both real-time rendering engines that can process dynamic images in real time and perform 3D rendering. The plug-in 302 is pre-installed in the rendering engine. The LLM 303 is configured on a server. The rendering engine 301 can call the LLM 303 through a pre-configured interface to realize communication between the rendering engine 301 and the LLM 303. The asset repository may be provided by the plug-in, or the asset repository may be a local asset repository self-defined by the user to generate a 3D scene with a user's personal style. The asset repository in FIG. 3 is displayed in the local asset repository 304.

[0033] LLM is a hot topic in the field of artificial intelligence in recent years. LLM is a pre-trained language model that can learn rich linguistic and world knowledge by pre-training on a large amount of text data, thereby achieving amazing results in various natural language processing (NLP) tasks. Wenxin Yiyin and ChatGPT are all applications developed based on LLM, which can generate fluent, logical, and creative text content and can also naturally interact with humans. Specifically, the large-scale model may be a Generative Pre-trained Transformer (GPT) model based on Transformer, an Enhanced Representation through Knowledge Integration (ERNIE) model based on knowledge integration, etc.

[0034] The asset repository includes assets, scene attribute information of assets, and material information of assets. Each information corresponds to a folder, which can be called an asset folder, a scene attribute folder, and a material folder. Each folder may be in a hierarchical format, for example, for the asset folder, the first level is the model, the second level includes vegetation, rocks, architecture, tools, etc., and the third level corresponding to vegetation includes trees, flowers, etc. For the scene attribute folder, the first level includes light, weather, screen style, asset position, etc., the second level corresponding to light includes light intensity, light color, light position, etc., and the second level corresponding to weather includes sunny, rainy, snow, etc. For the material folder, material adjustable parameters such as color, material attribute, snow / moss height, etc. can be recorded.

[0035] Taking the local asset repository as an example, the above various information (assets, material information, scene attribute information) is created in advance by the user. Scene attribute information can be created using an asset set tool, which can be an asset management tool or a level animation sequence tool, such as the Variant Manager tool and the LevelSequence tool. The Variant Manager tool and the LevelSequence tool are both asset set tools in UE, the Variant Manager tool is a set file that combines asset positions, material balls, various scene attributes, etc., and the LevelSequence tool can realize dynamic switching of assets / scene attributes by setting key frames.

[0036] Additionally, asset related information in the local asset repository may be recorded in an asset repository list, which may be referred to as candidate information.

[0037] The rendering engine provides an interactive interface to a user, which includes an input box in which the user can input descriptive information of a target 3D scene, for example, "I want to generate a scene of playing in a forest in northern China in the winter season, and in the scene there is a dirt road and a house made of piles, white smoke coming out of a chimney on the roof of the house, and warm yellow lights are on in the windows of the house, which is cool in color and realistic."

[0038] The rendering engine calls the LLM through a preset interface, and uses the LLM to perform extraction processing on the above-mentioned description information to obtain label information. For example, based on the above-mentioned description information, the extracted label information includes: a scene, winter season, northern China, forest, game, a dirt road, a house made of piles, a chimney on the roof of the house, white smoke coming out of the chimney, and a warm yellow light in the window of the house, with a cold color tone and realistic.

[0039] In this embodiment, by using LLM to extract label information in description information, more concise and accurate label information can be obtained based on the description information, which further improves the accuracy of the target asset set obtained based on the label information, and improves the accuracy and efficiency of generating 3D scenes.

[0040] After obtaining the LLM label information, the label information can be matched with the candidate information (candidate asset information, candidate material information, and candidate scene attribute information) in the asset repository list to obtain the target information (target asset information, target material information, and target scene attribute information) that matches the label information, and then the target information is sent to the plug-in.

[0041] The plug-in obtains a target asset set (target assets and corresponding target material information, and target scene attribute information) based on the target information, and generates a target 3D scene based on the target asset set.

[0042] The rendering engine renders the target 3D scene generated by the plug-in and then displays it to the user.

[0043] Combining the above application scenarios, the present disclosure further provides the following embodiments.

[0044] 4 is a schematic diagram according to a second embodiment of the present disclosure. This embodiment provides a method for generating a 3D scene based on a large-scale model, the method including the following steps: 401, generating extraction operation presentation information of an LLM based on description information of a target 3D scene, and using the LLM to process the description information based on the extraction operation presentation information to obtain label information in the description information.

[0045] For example, a user inputs description information, e.g., "I want to generate a scene of playing in the forest in northern China in winter season, and in the scene there is a dirt road and a house made of piles, white smoke coming out of the chimney on the roof of the house, and warm yellow lights are on in the windows of the house, in cool tones, and realistic," through an interface provided by the rendering engine. The rendering engine calls the LLM through a preset interface and uses the LLM to extract label information in the description information, e.g., the label information includes: one scene, winter season, northern China, forest, for playing, one dirt road, a house made of piles, a chimney on the roof of the house, white smoke coming out of the chimney, and warm yellow lights are on in the windows of the house, in cool tones, and realistic.

[0046] Specifically, the rendering engine can send a prompt to the LLM, which can be called an extraction operation prompt, and the extraction operation prompt includes description information and further includes instruction information, which is used to instruct the LLM to perform an information extraction operation, and the specific content of the instruction information can be set. The LLM can extract label information in the description information based on the extraction operation prompt.

[0047] After the LLM obtains the label information, it can further display the label information to the user.

[0048] In addition, the LLM can further determine the importance of each piece of label information and display each piece of label information in order based on its importance.

[0049] In this embodiment, by using LLM to extract label information in description information, more concise and accurate label information can be obtained based on the description information, which further improves the accuracy of the target asset set obtained based on the label information, and improves the accuracy and efficiency of generating 3D scenes.

[0050] 402. Generate query operation presentation information of an LLM based on the label information, and use the LLM to match the label information with a plurality of pre-recorded candidate information based on the query operation presentation information to obtain target information of a target asset set.

[0051] The multiple candidate information can be divided into three types: the first type is asset information, such as vegetation-northern-trees and ground-forest-land, architecture-house-tree, particles-smoke-chimney, etc.; the second type is scene attribute information, such as season-winter season, layout-forest, color tone-cool color tone, style-realistic, light-window; and the third type is material information, such as snow height, color hue, normal intensity, roughness intensity, etc. Taking the above asset information and scene attribute information as an example to be displayed in a hierarchical format, for example, for vegetation, the first layer is "vegetation", the second layer is north, the third layer is "trees and ground", and so on until the last layer is recorded, for example, "vegetation-north-trees and ground-forest-land" as mentioned above; similarly, assuming that for "vegetation" in the first layer, the second layer further includes vegetation information of "south", the related asset information can be displayed as vegetation-south-....

[0052] Multiple candidate information can be recorded in the asset repository list, and the LLM can obtain target information that matches the label information based on the pre-learned correspondence. For example, the label information "winter season" matches information such as "winter", "snow", "ice", and "cold" in the asset repository.

[0053] In this embodiment, the label information is matched with pre-recorded candidate information to obtain target information, so that the target information can be obtained simply and efficiently, and processing efficiency can be improved.

[0054] 403, using a pre-configured plug-in to obtain a set of target assets in a user self-defined local asset repository based on the target information.

[0055] After obtaining the target information, the LLM can send the target information to a plug-in pre-installed in the rendering engine, which can obtain a target asset set in a local asset repository based on the target information.

[0056] In this embodiment, a target asset set is obtained in a user-defined local asset repository, and the user can set candidate assets and their related information according to his or her own needs, and a personalized target 3D scene can be generated.

[0057] 404, using a pre-configured plug-in to generate a target 3D scene based on the target asset set.

[0058] Among them, the target asset set includes target assets, corresponding target material information, and target scene attribute information. Therefore, the plug-in can directly generate a target 3D scene based on this information, and the user does not need to set attributes such as asset position, which facilitates user operation.

[0059] Furthermore, the plug-in can further generate an initial 3D scene based on the target asset set, and adjust the initial 3D scene based on scene feature information in the label information to generate the target 3D scene.

[0060] For example, the above-mentioned label information includes "for games", and the information is scene function information, and the plug-in can preset operations corresponding to each function, and then adjust the initial 3D scene according to the preset operations. For example, according to the preset adjustment rules for games, the initial 3D scene can be adjusted in aspects such as the collective Level of Detail (LOD), light, resolution, rendering quality, etc. to obtain a target 3D scene.

[0061] In this embodiment, the initial 3D scene is adjusted according to the scene feature information to obtain the target 3D scene, so that the precision of the target 3D scene can be improved and the rendering effect can be improved.

[0062] 405, using a rendering engine to render the target three-dimensional scene.

[0063] After generating the target 3D scene using the plug-in, the rendering engine renders the target 3D scene and displays the rendering results to the user.

[0064] In addition, after the user obtains the rendering result, the user can modify the target 3D scene as necessary, either based on the label information or directly modify the asset or scene attribute information (e.g., position) in the target 3D scene.

[0065] Specifically, the label information can be displayed to a user, label information modified by the user can be obtained, a modified target asset set can be obtained based on the modified label information, and a modified target 3D scene can be generated based on the modified target asset set.

[0066] For example, if the label information above includes "winter", the user can modify the label information to "summer", and then the system can regenerate the summer-related scenes.

[0067] In this embodiment, the target 3D scene is modified based on the modified label information, eliminating the need for the user to perform specific modification operations on the scene, simplifying user operations and improving processing efficiency.

[0068] It is also possible to display the target 3D scene to a user and generate a modified target 3D scene based on modification instructions of the user.

[0069] Wherein, the modification command is, for example, to modify a model in the target 3D scene or to modify the position of the model, etc., and the modified target 3D scene can be obtained according to the modification command. For example, if there is a round table in the target 3D scene, the user can directly modify the round table into a square table in the scene.

[0070] In this embodiment, the target 3D scene is modified according to the modification command to obtain a personalized modified target 3D scene as required by the user.

[0071] 5 is a schematic diagram of a third embodiment of the present disclosure. This embodiment provides a large-scale model-based 3D scene generating device, as shown in FIG. 5, the device 500 includes a processing module 501, an acquisition module 502, and a generating module 503.

[0072] The processing module 501 is used to process description information of a target 3D scene to obtain label information in the description information, the acquisition module 502 is used to generate query operation presentation information of an LLM based on the label information, and use the LLM to obtain a target asset set matching the label information based on the query operation presentation information, the target asset set including target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets, and the generation module 503 is used to generate the target 3D scene based on the target asset set.

[0073] In this embodiment, the target asset set not only includes the target assets, but also includes scene attribute information and material information of the target assets, which is equivalent to obtaining the target assets, their materials, and scene attribute information such as their position and size at once. Compared to the method of obtaining only the target assets, a 3D scene can be generated more simply and efficiently.

[0074] In some embodiments, the processing module 501 is further used to generate extraction operation presentation information of the LLM based on the description information, and process the description information based on the extraction operation presentation information using the LLM to obtain the label information.

[0075] In this embodiment, by using LLM to extract label information in description information, more concise and accurate label information can be obtained based on the description information, which further improves the accuracy of the target asset set obtained based on the label information, and improves the accuracy and efficiency of generating 3D scenes.

[0076] In some embodiments, the acquisition module 502 is further used to match the label information with multiple pre-recorded candidate information based on the query operation presentation information using the LLM to acquire target information of the target asset set, and acquire the target asset set based on the target information.

[0077] In this embodiment, the label information is matched with pre-recorded candidate information to obtain target information, so that the target information can be obtained simply and efficiently, and processing efficiency can be improved.

[0078] In some embodiments, the acquisition module 502 is further used to acquire the set of target assets in a user self-defined local asset repository based on the target information.

[0079] In this embodiment, a target asset set is obtained in a user-defined local asset repository, and the user can set candidate assets and their related information according to his or her own needs, and a personalized target 3D scene can be generated.

[0080] In some embodiments, the generation module 503 is further used to generate an initial 3D scene based on the target asset set, and adjust the initial 3D scene based on scene feature information in the label information to generate the target 3D scene.

[0081] In this embodiment, the initial 3D scene is adjusted according to the scene feature information to obtain the target 3D scene, so that the precision of the target 3D scene can be improved and the rendering effect can be improved.

[0082] In some embodiments, the device 500 includes: The method further includes a first modification module for displaying the label information to a user, obtaining label information modified by the user, obtaining a modified target asset set based on the modified label information, and generating a modified target 3D scene based on the modified target asset set.

[0083] In this embodiment, the target 3D scene is modified based on the modified label information, eliminating the need for the user to perform specific modification operations on the scene, simplifying user operations and improving processing efficiency.

[0084] In some embodiments, the device 500 includes: The system further includes a second modification module for displaying the target three-dimensional scene to a user and generating a modified target three-dimensional scene based on modification instructions of the user.

[0085] In this embodiment, the target 3D scene is modified according to the modification command to obtain a personalized modified target 3D scene as required by the user.

[0086] It should be understood that the embodiments of the present disclosure may mutually refer to the same or similar content in different embodiments.

[0087] It will be understood that the terms "first", "second", etc. in the embodiments of the present disclosure are used only for distinction and do not represent importance, timing, etc.

[0088] In the technical solution disclosed herein, the collection, storage, use, processing, transmission, provision, disclosure and other processing of relevant user personal information all comply with relevant laws and regulations and do not violate public order and morals.

[0089] According to an embodiment of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium, and a computer program product.

[0090] As shown in FIG. 6, a block diagram of an electronic device 600 for implementing an embodiment of the present disclosure is shown. The electronic device is intended to represent various types of digital computers, such as laptop computers, desktop computers, workstations, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various types of mobile devices, such as personal digital processing, mobile phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure as described herein and / or claimed.

[0091] 6, the device 600 includes a computing unit 601, which can perform various appropriate operations and processes based on a computer program stored in a read-only memory (ROM) 602 or loaded from a storage unit 608 into a random access memory (RAM) 603. The RAM 603 can also store various programs and data required for the device 600 to operate. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0092] Several components in the electronic device 600 are connected to an I / O interface 605, including input units 606 such as a keyboard, mouse, etc., output units 607 such as various types of displays, speakers, etc., storage units 608 such as a disk, optical disk, etc., and communication units 609 such as a network card, modem, wireless communication transceiver, etc. The communication units 609 enable the device 600 to exchange information / data with other devices via a computer network, such as the Internet, and / or various telecommunication networks.

[0093] The computing unit 601 is a general-purpose and / or dedicated processing component with various processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, computing units that execute various machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 executes the various methods and processes described above, such as the method for generating a three-dimensional scene based on a large-scale model. For example, in some embodiments, the method for generating a three-dimensional scene based on a large-scale model can be realized as a computer software program tangibly included in a machine-readable medium, such as the storage unit 608. In some embodiments, some or all of the computer program is loaded and / or installed in the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the method for generating a three-dimensional scene based on a large-scale model can be executed. Alternatively, in other embodiments, the computation unit 601 may be configured to execute the method for generating a three-dimensional scene based on a large-scale model via any other suitable manner (eg, by firmware).

[0094] Various implementations of the systems and techniques described herein may be realized in digital electronic circuitry systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system on a chip (SOCs), programming programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include being implemented in one or more computer programs that may be executed and / or interpreted by a programmable system that includes at least one programmable processor, which may be an application specific or general purpose programmable processor, and that may receive data and instructions from and transmit data and instructions to a storage system, at least one input device, and at least one output device.

[0095] The program codes for implementing the methods of the present disclosure can be written using any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general purpose computer, a special purpose computer, or other programmable data processing apparatus such that the functions / operations specified in the flowcharts and / or block diagrams are implemented when the program codes are executed by the processor or controller. The program codes can be executed entirely on the machine, partially on the machine, partially on the machine as a separate software package and partially on a remote machine, or entirely on a remote machine or server.

[0096] In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain or store a program for use with or in combination with an instruction execution system, device, or apparatus. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium includes, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or apparatus, or any suitable combination of the above. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.

[0097] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user, and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other types of devices can also be used to provide interaction with a user, for example, feedback provided to the user can be any form of sensing feedback (e.g., visual feedback, auditory feedback, or haptic feedback) and can receive input from the user in any form (including acoustic input, speech input, and haptic input).

[0098] The systems and techniques described herein may be implemented in a computing system including a back-end component (e.g., a data server), or a middleware component (e.g., an application server), or a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user interacts with the implementation of the systems and techniques described herein), or any combination of such back-end, middleware, and front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), the Internet, and a blockchain network.

[0099] The computer system may include a client and a server. The client and the server are generally remote from each other and usually interact with each other via a communication network. The relationship between the client and the server is generated by a computer program that runs on a corresponding computer and has a client-server relationship with each other. The server may be a cloud server, also called cloud computing or cloud host, which is a host product in a cloud computing service system and solves the defects of high management difficulty and low business scalability that exist in traditional physical hosts and VPS services (abbreviated as "Virtual Private Server", or "VPS"). The server may be a server of a distributed system or a server that combines blockchain.

[0100] It should be understood that steps can be rearranged, added, or deleted using various types of flows shown above. For example, each step described in the present disclosure can be performed in parallel, sequentially, or in a different order, but is not limited herein as long as the technical solution disclosed in the present disclosure can achieve the desired results.

[0101] The above specific implementation methods do not constitute a limitation of the protection scope of the present disclosure. Those skilled in the art can make various modifications, combinations, subcombinations, and substitutions based on design requirements and other factors. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present disclosure should be included in the protection scope of the present disclosure.

Claims

1. A method for generating a three-dimensional scene based on a large scale model, comprising the steps of: processing a description of a target 3D scene to obtain label information in the description; generating query operation presentation information of an LLM according to the label information, and obtaining a target asset set matching the label information according to the query operation presentation information using the LLM, the target asset set including target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets; generating the target three-dimensional scene based on the target asset set. A method for generating 3D scenes based on large-scale models.

2. The step of processing description information of the target 3D scene to obtain label information in the description information includes: generating extraction operation presentation information of the LLM based on the description information; and processing the description information based on the extraction operation presentation information using the LLM to obtain the label information. The method for generating a three-dimensional scene based on a large scale model according to claim 1.

3. The step of obtaining a target asset set matching the label information based on the query operation presentation information using the LLM includes: Using the LLM, based on the query operation presentation information, matching the label information with a plurality of pre-recorded candidate information to obtain target information of the target asset set; obtaining the target asset set based on the target information. The method for generating a three-dimensional scene based on a large scale model according to claim 1.

4. The step of obtaining the target asset set based on the target information includes: obtaining the target asset set in a user-defined local asset repository based on the target information; The method for generating a 3D scene based on a large scale model according to claim 3.

5. generating the target three-dimensional scene based on the target asset set, generating an initial three-dimensional scene based on the target asset set; and adjusting the initial 3D scene based on scene feature information in the label information to generate the target 3D scene. The method for generating a three-dimensional scene based on a large scale model according to claim 1.

6. displaying the label information to a user; obtaining label information modified by the user; obtaining a modified target asset set based on the modified label information; and generating a modified target 3D scene based on the modified target asset set. A method for generating a three-dimensional scene based on a large-scale model according to any one of claims 1 to 5.

7. displaying the target three-dimensional scene to a user; generating a modified target three-dimensional scene based on the user's modification instructions. A method for generating a three-dimensional scene based on a large-scale model according to any one of claims 1 to 5.

8. An apparatus for generating a three-dimensional scene based on a large-scale model, comprising: a processing module for processing a description of a target 3D scene to obtain label information in the description; an acquisition module for generating query operation presentation information of an LLM according to the label information, and using the LLM to acquire a target asset set matching the label information according to the query operation presentation information, the target asset set including target assets in the target 3D scene, target material information of the target assets, and target scene attribute information of the target assets; a generation module for generating the target three-dimensional scene based on the target asset set. A device for generating 3D scenes based on large-scale models.

9. The processing module further comprises: generating extraction operation presentation information of the LLM based on the description information; and processing the description information based on the extraction operation presentation information using the LLM to obtain the label information; The apparatus for generating a three-dimensional scene based on a large scale model according to claim 8.

10. The acquisition module further comprises: Using the LLM, based on the query operation presentation information, match the label information with a plurality of pre-recorded candidate information to obtain target information of the target asset set; used to obtain the target asset set based on the target information. The apparatus for generating a three-dimensional scene based on a large scale model according to claim 8.

11. The acquisition module further comprises: used to obtain the target asset set in a user self-defined local asset repository based on the target information; The apparatus for generating a three-dimensional scene based on a large scale model according to claim 10.

12. The generating module further comprises: generating an initial three-dimensional scene based on the target asset set; and adjusting the initial 3D scene based on scene feature information in the label information to generate the target 3D scene. The apparatus for generating a three-dimensional scene based on a large scale model according to claim 8.

13. a first modification module for displaying the label information to a user, obtaining label information modified by the user, obtaining a modified target asset set based on the modified label information, and generating a modified target three-dimensional scene based on the modified target asset set. An apparatus for generating a three-dimensional scene based on a large-scale model according to any one of claims 8 to 12.

14. a second modification module for displaying the target three-dimensional scene to a user and generating a modified target three-dimensional scene based on modification instructions of the user. An apparatus for generating a three-dimensional scene based on a large-scale model according to any one of claims 8 to 12.

15. An electronic device, At least one processor; a memory in communication with the at least one processor; The memory stores instructions executable by the at least one processor, the instructions, when executed by the at least one processor, causing the at least one processor to perform a method according to any one of claims 1 to 5. electronic equipment.

16. A non-transitory computer-readable storage medium having computer instructions stored thereon, comprising: The computer instructions cause the computer to carry out a method according to any one of claims 1 to 5. A non-transitory computer-readable storage medium.

17. A computer program comprising: A computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5. Computer program.

Images