Two-dimensional engineering drawing scene conversion method and device and storage medium

Abstract

The invention relates to a two-dimensional engineering drawing scene conversion method and device and a storage medium, and is applied to the technical field of digital twinning, and the method comprises the steps: converting a CAD two-dimensional engineering drawing into a DXF file, uploading the DXF file to a cloud, recognizing equipment corresponding to the DXF file through a preset industrial large model in the cloud, obtaining a standard three-dimensional model corresponding to each kind of equipment, and storing the standard three-dimensional model in the cloud; carrying out layout on the standard three-dimensional model of each type of equipment to obtain a three-dimensional model scene; the method comprises the following steps: acquiring text description of motion of each device in a three-dimensional model scene input by a user, extracting motion keywords of each device from a pre-trained Transform model, generating an action sequence based on the motion keywords, and generating a three-dimensional demonstration animation according to the action sequence based on a pre-trained GAN generative adversarial network; the method realizes conversion from the two-dimensional drawing to the three-dimensional animation, generates the animation in real time, has the advantages of low cost and short time consumption, and does not need to worry about scheme leakage caused by delivery to an animation company.

Description

Technical Field

[0001] The present invention relates to the field of digital twin technology, and in particular to a two-dimensional engineering drawing scene conversion method, device and storage medium. Background Art

[0002] For the warehousing and logistics industry, computer-aided design (CAD) is a widely used modeling technology that can greatly improve the work efficiency of designers and realize the overall design of the project. After the design unit designs the plan for the client, it needs to explain, demonstrate and verify the plan to the client. In the process of explanation and demonstration, it is mostly presented through two-dimensional CAD drawings. However, this display method cannot intuitively let customers understand the overall picture of the plan, and cannot fully reflect the effect of the specific implementation plan. The communication effect is poor, and it is even more difficult for non-professionals to understand.

[0003] In order to improve the communication effect between the two parties, the design unit will find some animation process production companies to make the scheme effect diagram or animation according to the CAD scheme engineering drawings, and provide them to Party A in a more intuitive way to improve the communication efficiency of both parties;

[0004] Presenting the plan in the form of renderings or animations will give people a more three-dimensional and intuitive understanding of the overall plan. However, for the design unit, this method will undoubtedly add extra costs if an animation company is asked to produce it. Once the plan is modified, the renderings or animations will need to be re-produced, which will significantly increase the time and money costs. Furthermore, if the reserved period is too short, it will be difficult for the animation production company to complete it in a short period of time, which will cause delays in communication between the two parties. More seriously, it will cause the project to fail. At the same time, the dissemination of project plan drawings will also lead to the leakage of the unit's technical plans and confidential information. Summary of the invention

[0005] In view of this, the purpose of the present invention is to provide a two-dimensional engineering drawing scene conversion method, device and storage medium to solve the problems in the prior art that the cost of producing renderings or animations by animation companies is high, the production time of animation companies is long, and there is a risk of leakage of scheme confidentiality when sending drawings to animation companies.

[0006] According to a first aspect of an embodiment of the present invention, a method for converting a two-dimensional engineering drawing scene is provided, the method comprising:

[0007] Obtaining a CAD two-dimensional engineering drawing to be converted, and saving the CAD engineering drawing to be converted as a DWG format file;

[0008] Output the DWG format file into a DXF file and upload it to a cloud server, identify the DXF file through the industrial model preset in the cloud server, and obtain a standard 3D model corresponding to the DXF file of each device;

[0009] The standard three-dimensional model of each device is laid out according to the CAD two-dimensional engineering drawing to obtain a three-dimensional model scene;

[0010] Obtaining a text description of the motion of each device in the 3D model scene, inputting the text description into a pre-trained Transformer model, and extracting motion keywords for each device through the pre-trained Transformer model;

[0011] Creating an action sequence in the three-dimensional model scene according to the motion keywords of each device;

[0012] Based on the pre-trained GAN generative adversarial network, image sequence frames of each device are generated according to the action sequence, and the image sequence frames of all devices are arranged into a continuous video stream on the timeline to obtain an animated demonstration of the three-dimensional model scene.

[0013] Preferably,

[0014] The identifying of the DXF file by using the industrial model preset in the cloud server includes:

[0015] Convert the standard blocks of each device in the CAD software into DXF files in advance and upload them to the cloud server, and perform learning and recognition training on the DXF files of each device through the industrial large model preset in the cloud server;

[0016] After the cloud server obtains the DXF file of each device in the CAD two-dimensional engineering drawing to be converted, the device corresponding to the DXF file is identified through the pre-trained industrial large model.

[0017] Preferably,

[0018] The step of obtaining the standard 3D model corresponding to the DXF file of each device includes:

[0019] Converting the standard 3D model of each device into an FBX file in advance and uploading it to the cloud server;

[0020] When the cloud server obtains the device corresponding to the DXF file of each device in the CAD two-dimensional engineering drawing to be converted;

[0021] Obtain the FBX file corresponding to each device, and render a standard 3D model of each device based on the FBX file of each device.

[0022] Preferably,

[0023] The step of converting the standard 3D model of each device into an FBX file in advance and uploading the file to the cloud server further includes:

[0024] Convert the standard 3D models of each device of different device manufacturers into FBX files and upload them to the cloud server;

[0025] The method of obtaining the FBX file corresponding to each device includes:

[0026] Select the device manufacturer and retrieve the FBX file corresponding to each device of the corresponding device manufacturer.

[0027] Preferably,

[0028] The obtaining of the standard three-dimensional model of each device further comprises:

[0029] The height of the standard 3D model for each device is defined, and the standard 3D model is stretched in the vertical direction based on the defined height of the standard 3D model.

[0030] Preferably,

[0031] The movement keywords of each device include: the movement direction of the device, the starting and ending points of the device, and the duration of the device movement process.

[0032] According to a second aspect of an embodiment of the present invention, a two-dimensional engineering drawing scene conversion device is provided, the device comprising:

[0033] DWG format conversion module: used to obtain the CAD two-dimensional engineering drawing to be converted, and save the CAD engineering drawing to be converted as a DWG format file;

[0034] 3D model conversion module: used to output the DWG format file into a DXF file and upload it to the cloud server, identify the DXF file through the industrial model preset in the cloud server, and obtain the standard 3D model corresponding to the DXF file of each device;

[0035] A three-dimensional scene acquisition module is used to lay out the standard three-dimensional model of each device according to the CAD two-dimensional engineering drawing to obtain a three-dimensional model scene;

[0036] Motion keyword acquisition module: used to obtain text descriptions of the motion of each device in the 3D model scene, input the text descriptions into a pre-trained Transformer model, and extract motion keywords for each device through the pre-trained Transformer model;

[0037] An action sequence generation module: used for creating an action sequence in the three-dimensional model scene according to the motion keywords of each device;

[0038] 3D animation generation module: used to generate an adversarial network based on a pre-trained GAN, generate image sequence frames for each device according to the action sequence, arrange the image sequence frames of all devices into a continuous video stream on the timeline, and obtain an animation demonstration of the 3D model scene.

[0039] According to a third aspect of an embodiment of the present invention, there is provided a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented.

[0040] The technical solution provided by the embodiments of the present invention may have the following beneficial effects:

[0041] This application converts CAD two-dimensional engineering drawings into DXF files and uploads them to the cloud, identifies the devices corresponding to the DXF files through the industrial large models preset in the cloud, obtains the standard three-dimensional models corresponding to each device, and obtains the three-dimensional model scene by laying out the standard three-dimensional models of each device; obtains the text description of the movement of each device in the three-dimensional model scene input by the user, extracts the movement keywords of each device from the pre-trained Transformer model, generates an action sequence based on the movement keywords, and generates a three-dimensional demonstration animation based on the action sequence based on the pre-trained GAN generative adversarial network; realizes the conversion of two-dimensional drawings to three-dimensional animations, and generates animations in real time, which has the advantages of low cost and short time consumption, and there is no need to worry about the leakage of the plan due to handing it over to the animation company.

[0042] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0044] Figure 1 is a flow chart of a two-dimensional engineering drawing scene conversion method according to an exemplary embodiment;

[0045] Figure 2 is a system schematic diagram of a two-dimensional engineering drawing scene conversion device according to another exemplary embodiment;

[0046] In the attached figure: 1-DWG format conversion module, 2-3D model conversion module, 3-3D scene acquisition module, 4-motion keyword acquisition module, 5-action sequence generation module, 6-3D animation generation module. DETAILED DESCRIPTION

[0047] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0048] Embodiment 1

[0049] Figure 1 is a flow chart of a two-dimensional engineering drawing scene conversion method according to an exemplary embodiment. Figure 1 As shown, the method includes:

[0050] S1, obtaining a CAD two-dimensional engineering drawing to be converted, and saving the CAD engineering drawing to be converted as a DWG format file;

[0051] S2, outputting the DWG format file into a DXF file and uploading it to a cloud server, identifying the DXF file through an industrial model preset in the cloud server, and obtaining a standard 3D model corresponding to the DXF file of each device;

[0052] S3, laying out the standard three-dimensional model of each device according to the CAD two-dimensional engineering drawing to obtain a three-dimensional model scene;

[0053] S4, obtaining a text description of the movement of each device in the three-dimensional model scene, inputting the text description into a pre-trained Transformer model, and extracting movement keywords of each device through the pre-trained Transformer model;

[0054] S5, creating an action sequence in the three-dimensional model scene according to the motion keywords of each device;

[0055] S6, based on the pre-trained GAN generative adversarial network, generating image sequence frames for each device according to the action sequence, arranging the image sequence frames of all devices into a continuous video stream on the time axis, and obtaining an animation demonstration of the three-dimensional model scene;

[0056] It is understandable that in mature warehousing and logistics companies, there are corresponding complete CAD block libraries to represent equipment types. When designing schemes in CAD, scheme design and planning engineers only need to find the standard blocks of the corresponding equipment and copy and paste them, and then lengthen, enlarge or reduce them according to actual needs. Therefore, a complete factory workshop process plan is composed of countless CAD blocks. Therefore, the CAD standard blocks of each equipment must be converted into code form for the logistics industry large model to learn;

[0057] In the CAD software, the standard blocks of each equipment can be exported into DXF files. The DXF data format is an open CAD standard that can well record geometric shapes, layers, text annotations, dimension attributes and other element attributes. The DXF files of various equipment are stored in the cloud server for learning, training and calling of large models of logistics industry.

[0058] Since the CAD plan drawing is a two-dimensional plane, it only records the length and width of the equipment, and does not reflect the height of the equipment. However, the height of the equipment is also one of the important aspects of the plan. In addition to the height of the equipment, the CAD block of the equipment can only show an approximate height because it is in a standard form. There are slight differences in the equipment of each manufacturer, so the equipment manufacturer factor also needs to be considered;

[0059] This application saves the standard 3D models of each type of equipment of each manufacturer and stores them in FBX format to reduce memory usage, and uploads these FBX format file data to the cloud server for generation and calling. The FBX file can well include the materials, textures, animations, and skeleton information in the 3D model. When stored in the cloud server, it is also convenient for format conversion and subsequent loading in WebGL.

[0060] When the engineer designs the complete solution layout in the CAD software, each type of equipment is processed into blocks to form a complete DWG format file, the DWG file is exported to a DXF file, and then the DXF file of the scene 2D drawing is uploaded to the cloud server as input information;

[0061] The cloud server recognizes the content of the DXF file based on the preset logistics industry model (since the logistics industry model has been trained in advance through the equipment DXF files in the CAD image library, the logistics industry model can recognize the corresponding equipment according to the DXF file). After recognizing the equipment corresponding to the DXF file, the cloud server finds the FBX file of the corresponding equipment based on the equipment information in the successfully parsed DXF file scene data (the above description pre-selects the FBX files and standard 3D models corresponding to each type of equipment produced by different manufacturers and uploads them to the cloud server), and renders the 3D model scene of the corresponding equipment based on the FBX file; note that because the 2D CAD drawing cannot reflect the equipment manufacturer and equipment height, after rendering the 3D model scene, the design engineer needs to select the equipment manufacturer according to the design situation and actual user needs. After the selection is successful, the cloud server retrieves the corresponding manufacturer's equipment FBX file to generate a 3D model;

[0062] Although the standard 3D model itself contains the height, for the logistics industry, the height of some equipment needs to be freely adjusted, such as conveyors. Assume that the height of the standard 3D model of the conveyor corresponding to the manufacturer is 3m, but the actual requirement is 5m, so the height needs to be adjusted according to the needs based on the standard 3D model;

[0063] Therefore, by defining the height of the device, the 3D model of the device is stretched in the vertical direction, and finally the 3D model layout and rendering of the entire scene are completed. For projects that only require renderings, only the first part of the 3D model scene rendering is completed;

[0064] After the 3D model scene of the solution is built, it is still a static representation, which can give people a more intuitive overall visual effect and an understanding of the project as a whole, but there is no clear description of the entire solution process and work flow. This requires the production of process animation for a clearer expression;

[0065] Because the overall 3D scene has been constructed, in the field of logistics, the process of making animation is more about expressing the flow of equipment and goods and the movement of equipment, so design engineers can express it well through text without considering the influence of other factors;

[0066] The text content of the device movement input by the design engineer can be converted into the object movement in the scene. The more detailed and simple the text description of the design engineer is, such as defining the object's movement direction, starting and ending points, and the duration of the movement process, the more realistic the generated animation will be.

[0067] The animation generation part is first based on the Transformer architecture, which is a neural network architecture based on the self-attention mechanism. It can process all the position information in the input text at the same time, so that the model can capture the global context information, extract keywords from the input text based on semantic analysis, and focus on extracting the object, position, action and time sequence information.

[0068] In the Natural Language Processing (NLP) system, Transformer is a mainstream model and key technology that integrates the attention mechanism and neural network model. Transformer has the ability to vectorize or vectorize all the processed text and sentences to maximize the reflection of accurate meaning. The basic principles of the Transformer structure are:

[0069] Tokenization or patching: Whether processing text, images, or video data, the first step of the Transformer model is usually to convert the input data into a series of tokens or patches. For text, these tokens are usually words or parts of words;

[0070] Embedding Layer: After being converted to tokens, these tokens pass through an embedding layer. The role of the embedding layer is to convert each token into a fixed-size vector that the model can handle. For text tokens, this usually involves converting words into points in a high-dimensional space;

[0071] Self-Attention: Self-attention is at the heart of the Transformer. This mechanism enables the model to weigh the importance of each token with other tokens in the sequence, which means the model can learn how each token relates to other tokens, thereby better understanding the entire sequence.

[0072] Multi-Head Attention: In this stage, the model does not learn the relationship between input tokens from only one perspective, but learns from multiple different perspectives at the same time, which can help the model capture richer information in different subspaces;

[0073] Positional Encoding: Since the Transformer itself does not have the ability to process the order information of the sequence like the recurrent neural network (RNN), positional encoding needs to be added to provide the position information of each token in the sequence, which ensures that the model can take into account the order of tokens;

[0074] Encoder-Decoder Structure: Most Transformer models consist of two parts: an encoder and a decoder. The encoder is responsible for processing input tokens, while the decoder generates corresponding outputs based on the encoder's output. This structure is particularly suitable for tasks such as machine translation, where both the input and output are sequence data.

[0075] Secondly, we use computer vision and animation synthesis technology to create action sequences in the 3D model scene based on the extracted keyword information, and generate high-quality image sequence frames based on the created action sequences based on the GAN generative adversarial network; finally, we arrange all the generated image sequence frames into a continuous video stream on the timeline to ensure the smoothness of the generated video. If we need to add special effects such as background music and subtitles later, we need to import the generated video into professional video software such as PR for editing;

[0076] In the GAN production network, the Diffusion model is used. It is a generative model for image generation. It learns to generate new data by simulating the diffusion process commonly seen in nature. It mainly includes two key steps, namely, gradually confusing the data by adding (Gaussian) noise (diffusion process encoder), and then learning to reverse the noise process to reconstruct the sample (inverse diffusion process decoder);

[0077] Diffusion process Encoder: compresses the output of Transformer into a low-dimensional space. The compressed data will be full of noise for model training and learning;

[0078] Inverse diffusion process Decoder: Derived from the noise in reverse, gradually eliminating the noise to reverse the generated image, that is, to restore and create the compressed data to restore it to the high-definition pixel space;

[0079] Specifically: The operation mechanism of converting text description into video includes text processing, text encoding, video generation network, generator and discriminator training, etc.

[0080] Text processing: After the text is input, it is preprocessed and analyzed, the input text is segmented, the text is divided into single words and phrases, and the segmented structure is embedded, and each word and phrase is converted into a vector representation for subsequent text encoding;

[0081] Text encoding: The input text description is passed through a text encoder to convert the text into a vector representation. This vector is a mathematical representation in a high-dimensional space that can capture the semantic and grammatical information of the text description.

[0082] Video generation network: After the text encoder obtains the vector representation, it is input into the video generation network. The video generation network is usually composed of a generator and a discriminator. The generator is responsible for generating video content, and the discriminator is responsible for evaluating the authenticity of the generated video.

[0083] Generator generates video: The generator receives the vector representation output by the text encoder and generates the corresponding video content based on the vector representation. The generator generates corresponding video frames based on the input text description and combines them into complete video content. The output of the generator can be a static image, a dynamic video or a three-dimensional scene.

[0084] Training process: In the process of generating videos, the generator will continuously optimize its own parameters to make the generated video content closer to the real video. This optimization process is achieved through the game between the generator and the discriminator. The discriminator evaluates whether the video generated by the generator is real, and the generator adjusts the video content generated by itself based on the feedback. The training process usually includes five stages: First, the data preparation stage, collecting a large amount of training data containing text descriptions and corresponding video content to facilitate model training and learning; second, the model initialization stage, initializing the parameters of the video generation network, including the weights of the generator, discriminator or encoder and decoder; third, the generator training stage, by minimizing the difference between the video generated by the generator and the real video, the generator is trained to generate more realistic video content; fourth, the discriminator or encoder and decoder training stage, by minimizing the discriminator's ability to distinguish between the generated video and the real video, or minimizing the encoder and decoder The reconstruction error of the video content is trained, the discriminator or encoder and decoder; fifth, the model optimization stage, by adjusting the parameters of the model, the quality and realism of video generation are improved;

[0085] Generate video output: After the generator generates video content, it can be output as a video file or displayed on the screen in real time;

[0086] The whole process can be summarized into four steps. First, input text description: the user provides a text description for generating video content; second, text encoding: the input text description is segmented, word embedded and text encoded to obtain a text vector representation; third, generate video content: the text vector representation is input into the generator, and the generator generates the corresponding video according to the text description; fourth, video output: the generated video content is output to the user;

[0087] The Transformer structure recognizes the input text, converts the text into data, and generates detailed single-frame images through the diffusion model, which are then combined into a continuous video.

[0088] It is worth emphasizing that when the industrial model on the cloud server cannot parse the information of a certain device in the DXF file, the device will not appear on the display interface. When this happens, first check the CAD drawings involved, especially the block drawing of the device. If there is no abnormality, repeat the above operation again. If the device still cannot be displayed, it means that there is no model data of the device in the cloud model library. If the designer has a 3D model of the device or the designer can self-model the device, the designer can upload the 2D CAD block, DXF data and 3D model of the device to the cloud server, and learn the uploaded information through the logistics industrial model. After successful learning, try again;

[0089] However, when the designer does not have a 3D model of the device, he needs to first collect pictures of devices of the same style and brand as the device or similar devices. The more pictures from different angles, the better. These pictures are also uploaded to the cloud server. Then the design engineer makes a text description based on the specific style of the device. The description must include the name of the device and its length, width and height. By parsing the device name in the text content, the system can retrieve the photo of the device from the cloud server and generate a simple device model based on the GAN generative adversarial network based on the photo of the device. If the generated simple device model cannot meet the requirements, the designer needs to further accurately define the device through text description until the generated model meets the requirements.

[0090] Embodiment 2

[0091] Figure 2 is a system schematic diagram of a two-dimensional engineering drawing scene conversion device according to another exemplary embodiment, the device comprising:

[0092] DWG format conversion module 1: used to obtain the CAD two-dimensional engineering drawing to be converted, and save the CAD engineering drawing to be converted as a DWG format file;

[0093] 3D model conversion module 2: used to output the DWG format file into a DXF file and upload it to the cloud server, identify the DXF file through the industrial model preset in the cloud server, and obtain the standard 3D model corresponding to the DXF file of each device;

[0094] 3D scene acquisition module 3: used to lay out the standard 3D model of each device according to the CAD 2D engineering drawing to obtain a 3D model scene;

[0095] Motion keyword acquisition module 4: used to obtain text descriptions of the motion of each device in the 3D model scene, input the text descriptions into a pre-trained Transformer model, and extract motion keywords for each device through the pre-trained Transformer model;

[0096] Action sequence generation module 5: used for creating an action sequence in the three-dimensional model scene according to the motion keywords of each device;

[0097] 3D animation generation module 6: used to generate an adversarial network based on a pre-trained GAN, generate image sequence frames for each device according to the action sequence, and arrange the image sequence frames of all devices into a continuous video stream on the timeline to obtain an animation demonstration of the 3D model scene.

[0098] Embodiment three:

[0099] This embodiment provides a storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a main controller, each step in the above method is implemented;

[0100] It is understandable that the storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0101] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.

[0102] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.

[0103] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0104] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0105] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0106] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0107] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0108] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0109] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A two-dimensional engineering drawing scene conversion method, characterized in that: The method comprises: Obtaining a CAD two-dimensional engineering drawing to be converted, and saving the CAD engineering drawing to be converted as a DWG format file; Output the DWG format file into a DXF file and upload it to a cloud server, identify the DXF file through the industrial model preset in the cloud server, and obtain a standard 3D model corresponding to the DXF file of each device; The standard three-dimensional model of each device is laid out according to the CAD two-dimensional engineering drawing to obtain a three-dimensional model scene; Obtaining a text description of the motion of each device in the 3D model scene, inputting the text description into a pre-trained Transformer model, and extracting motion keywords for each device through the pre-trained Transformer model; Creating an action sequence in the three-dimensional model scene according to the motion keywords of each device; Based on the pre-trained GAN generative adversarial network, image sequence frames of each device are generated according to the action sequence, and the image sequence frames of all devices are arranged into a continuous video stream on the timeline to obtain an animated demonstration of the three-dimensional model scene.

2. The method according to claim 1, characterized in that: The identifying of the DXF file by using the industrial model preset in the cloud server includes: Convert the standard blocks of each device in the CAD software into DXF files in advance and upload them to the cloud server, and perform learning and recognition training on the DXF files of each device through the industrial large model preset in the cloud server; After the cloud server obtains the DXF file of each device in the CAD two-dimensional engineering drawing to be converted, the device corresponding to the DXF file is identified through the pre-trained industrial large model.

3. The method according to claim 2, characterized in that The step of obtaining the standard 3D model corresponding to the DXF file of each device includes: Converting the standard 3D model of each device into an FBX file in advance and uploading it to the cloud server; When the cloud server obtains the device corresponding to the DXF file of each device in the CAD two-dimensional engineering drawing to be converted; Obtain the FBX file corresponding to each device, and render a standard 3D model of each device based on the FBX file of each device.

4. The method according to claim 3, characterized in that The step of converting the standard 3D model of each device into an FBX file in advance and uploading the file to the cloud server further includes: Convert the standard 3D models of each device of different device manufacturers into FBX files and upload them to the cloud server; The method of obtaining the FBX file corresponding to each device includes: Select the device manufacturer and retrieve the FBX file corresponding to each device of the corresponding device manufacturer.

5. The method according to claim 4, characterized in that The obtaining of the standard three-dimensional model of each device further comprises: The height of the standard 3D model for each device is defined, and the standard 3D model is stretched in the vertical direction based on the defined height of the standard 3D model.

6. The method according to claim 5, characterized in that The movement keywords of each device include: the movement direction of the device, the starting and ending points of the device, and the duration of the device movement process.

7. A two-dimensional engineering drawing scene conversion device, characterized in that: The device comprises: DWG format conversion module: used to obtain the CAD two-dimensional engineering drawing to be converted, and save the CAD engineering drawing to be converted as a DWG format file; 3D model conversion module: used to output the DWG format file into a DXF file and upload it to the cloud server, identify the DXF file through the industrial model preset in the cloud server, and obtain the standard 3D model corresponding to the DXF file of each device; A three-dimensional scene acquisition module is used to lay out the standard three-dimensional model of each device according to the CAD two-dimensional engineering drawing to obtain a three-dimensional model scene; Motion keyword acquisition module: used to obtain text descriptions of the motion of each device in the 3D model scene, input the text descriptions into a pre-trained Transformer model, and extract motion keywords for each device through the pre-trained Transformer model; An action sequence generation module: used for creating an action sequence in the three-dimensional model scene according to the motion keywords of each device; 3D animation generation module: used to generate an adversarial network based on a pre-trained GAN, generate image sequence frames for each device according to the action sequence, arrange the image sequence frames of all devices into a continuous video stream on the timeline, and obtain an animation demonstration of the 3D model scene.

8. A storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by the main controller, each step in the two-dimensional engineering drawing scene conversion method as described in any one of claims 1-6 is implemented.

Images