77 results about "Model graph" patented technology

The invention discloses a civil engineering construction task automatic scheduling method and system. The method comprises the steps that a heterogeneous graph model representing construction task elements and the incidence relation of the construction task elements is constructed; processing the heterogeneous graph model by using a graph neural network to mine a dependency relationship between processes and extract risk association features related to uncertainty to form a construction state graph; constructing a Bayesian network based on the construction state atlas; probabilistic reasoning is carried out through a Bayesian network, the influence of uncertain factors is quantified, and key information representing risk conduction is fed back to the graph neural network so as to dynamically correct recognition of the graph neural network on a process dependency relationship; the dependency constraint output by the graph neural network and the risk quantification result output by the Bayesian network are fused, and a construction task scheduling scheme is generated; and synchronously updating the heterogeneous graph model, the graph neural network and the Bayesian network according to field construction feedback data to realize closed-loop iterative optimization of a scheduling scheme.

The application relates to the technical field of image processing, and provides a model material replacement method and device for a design drawing. The method comprises the following steps: determining a target model from each model of a design drawing according to a received material selection instruction; obtaining a target material map corresponding to the material selection instruction from each material map of the target model; and merging the target material map, a light map, a diffuse reflection map and a reflection refraction map of the design drawing to obtain a model image of the target model in the design drawing; wherein the design drawing is an image obtained through rendering; the light map, the diffuse reflection map and the reflection refraction map are each layer obtained after the design drawing is rendered. The model material replacement method for the design drawing provided by the application can reduce the calculation amount of the model in the design drawing when the material is replaced, and reduce the resources consumed when the model material is replaced.

This application relates to the field of simulation technology and discloses a method, apparatus, electronic device, and storage medium for simulation result analysis and evaluation. The method includes: in response to receiving any simulation result, determining whether the simulation result meets a performance target; when the simulation result meets the performance target, at least extracting the model diagram of the simulation model to which the simulation result belongs and at least one directly related scheme parameter, and obtaining at least one non-directly related scheme parameter related to the performance target and / or the simulation result, to determine the evaluation value and influence coefficient corresponding to each non-directly related scheme parameter; calculating a simulation evaluation score based at least on the evaluation values ​​and influence coefficients corresponding to all non-directly related scheme parameters, and then outputting an overall simulation evaluation; calling a simulation report template, and generating simulation optimization suggestions and an evaluation report based on the template, at least according to the overall simulation evaluation, the model diagram, and at least one directly related scheme parameter. This application facilitates the automated optimization and evaluation of simulation results.

The application provides a hierarchical automatic layout method and device for internal module graphs, comprising the following steps: constructing a preliminary shape of a model graph layout; scanning each node, if there is a nested relationship, taking the nested information as a subgraph of the current node; scanning each node of the nested information, if there is a nested relationship, taking the nested information as a subgraph of the current node, and constructing a preliminary shape of a nested graph layout; determining virtual connections between different node nested relationships; and correcting the preliminary shape of the model graph layout to obtain a final shape of the model layout. In the layout process, the ports and graph element nesting involved in the internal module graph are considered, and a hierarchical automatic layout method completely covering the internal module graph elements is realized.

The invention discloses a protein and small molecule binding structure prediction method, device, equipment and medium, and relates to the technical field of protein and small molecule binding prediction.The method comprises the steps that a first modeling graph structure and a second modeling graph structure after a protein block and a to-be-tested small molecule compound are modeled respectively are obtained; the first modeling graph structure and the second modeling graph structure are input into a TANKBind model, and a predicted protein small molecule compound is output through the TANKBind model; generating a first topological file and a second topological file corresponding to the protein structure file and the small molecule structure file of the protein small molecule compound; performing molecular dynamics simulation by using a preset molecular dynamics simulation tool according to a simulation file generated by the first topological file and the second topological file to obtain a molecular dynamics simulation track file; and analyzing the molecular dynamics simulation track file, the first topology file and the second topology file to obtain a simulation analysis result.

The application provides a large model inference optimization method, system, medium, terminal and program product for Ascend processors. The method comprises: comparing the automatically detected version with a built-in compatibility table; if the versions do not match, automatically loading a runtime patch or calling a backup library; if the versions match, processing a user request based on a preset hybrid scheduling mechanism; dynamically selecting a MindIE inference framework or a vLLM inference framework based on a runtime adaptation layer according to the user request parameters; converting large model weights into low-bit weights and compensating based on an error compensation matrix; identifying subgraphs and fusing operators for a model graph according to the computing characteristics of the Ascend processor, and rearranging intermediate tensors according to the 512B memory access alignment rule of the Ascend processor; performing pipelined overlap processing on the matrix multiplication calculation of the current token and the full aggregation operation based on an asynchronous execution flow; and performing inference and generating an inference result according to the user request. The application can improve the efficiency and reliability of large model inference.

The invention relates to an automobile automatic window falling method, system and device, and belongs to the technical field of automobile control. The water falling working condition is analyzed based on the real-time invasion speed of the water surface relative to an automobile door and the distance between the water surface and the lower edge of an automobile window, and an analysis result is obtained; the method comprises the following steps: acquiring environment characteristic data information around an automobile by using an outside-automobile look-around fisheye camera and a millimeter-wave radar, and performing abnormality judgment according to the environment characteristic data information around the automobile and a model graph loaded by the automobile to obtain a first abnormality judgment result or a second abnormality judgment result, and generating an emergency window descending scheme according to the first abnormity judgment result or the second abnormity judgment result. Through multi-sensor fusion and multi-condition judgment after the automobile falls into water, the window falling function can be provided when the automobile falls into water, whether obstacles exist around after the automobile falls into water to prevent opening of automobile windows or not is judged, passengers can be further prevented from being trapped in the automobile after the automobile falls into water, and great life safety hidden dangers brought to the passengers are avoided.

The present application relates to the technical field of hub polishing robots, and particularly relates to a hub polishing robot trajectory planning method based on 3D vision curve technology; by comparing an actual 3D curved surface graph with a preset 3D curved surface graph, a to-be-polished region model graph of each to-be-polished region in a target hub is generated; according to the to-be-polished region model graph of each to-be-polished region, optimal polishing parameters for polishing processing of each to-be-polished region in the target hub are obtained by pairing in a database; according to the optimal polishing parameters for polishing processing of each to-be-polished region in the target hub, optimal polishing trajectories of a polishing robot are planned; and the optimal polishing trajectories of the polishing robot and the optimal polishing parameters for polishing processing of each to-be-polished region in the target hub are sent to a control terminal. More accurate, self-adaptive, efficient and consistent polishing effects can be achieved, and production efficiency and product quality are improved.

The application aims to provide a network community class group and user unified modeling graph pre-training method, which comprises the following steps: constructing a heterogeneous graph model with hierarchical features by extracting semantic information and structural information of two different objects of community level and user level; in the pre-training process of the model, by introducing a self-supervised task that can adapt to different hierarchical objects, the model can represent the characteristics of different hierarchical objects and the mutual relationship between the hierarchical levels; in the pre-training process of the model, by designing a hierarchical iterative training method, the model can promote the mutual representation between different hierarchical levels, and complete various complex inference tasks with community and user as target objects.

The invention relates to the technical field of drawing recognition, in particular to an architectural drawing element intelligent recognition and labeling method based on graphic feature extraction, comprising: acquiring an architectural drawing file to be processed, the architectural drawing file to be processed comprising a two-dimensional CAD drawing and a three-dimensional building information model drawing; format analysis is conducted on the to-be-processed building drawing file, graph geometric data and layer description information are extracted, the graph geometric data comprise point coordinates, line coordinates and plane coordinates forming a building graph, and the layer description information comprises name information and layer priority information of building drawing layers; and building a building drawing feature database according to the graphic geometric data and the layer description information. According to the method, through deep fusion analysis of the graphic geometric data and the layer description information, and through the structured drawing feature database and the feature fusion module, various elements in the architectural drawing can be accurately identified and labeled, and the accuracy of drawing labeling can be effectively improved through the accurate identification.

The application relates to a low-altitude signal coverage prediction method and system based on spherical non-uniform sampling, wherein the method comprises the following steps: determining the region complexity of different regions according to a digital surface model graph, constructing a direction sampling weight function according to the region complexity and an antenna directional diagram gain; performing spherical non-uniform sampling on a unit sphere based on the direction sampling weight function to generate a ray direction set matched with the sampling weight; calculating the spherical solid angle corresponding to each sampling direction based on the ray direction set, calculating the transmission power corresponding to each ray according to the spherical solid angle and a base station antenna directional diagram; taking the ray direction set and the transmission power as inputs to perform ray tracking propagation calculation to obtain the received signal power of each sampling point at low altitude, and the low-altitude signal coverage prediction is completed. The problems of how to improve the efficiency and accuracy of low-altitude signal coverage prediction in the related art are solved.

1. Name of the product in this design: Response item for a graphical user interface of an intelligent dialogue system for electronic devices. 2. Intended use of this design: for use in an electronic device. 3. The key design features of this product are: the graphical user interface that uses non-dashed lines to represent the interface content. 4. The picture or photo that best illustrates the key design points: Design 1 front view. 5. Design 1 is designated as the basic design. 6. Purpose of the graphical user interface: The graphical user interface of this design is used as a whole for an intelligent dialogue system, and the protected part is used for the response items of the graphical user interface of the intelligent dialogue system. Intelligent dialogue systems are representations of intelligent agents. The response items of the intelligent dialogue system include cards for presenting feedback information from multiple different models, a control bar for switching between multiple cards, and interactive controls for the control bar. In Design 1, users can click on the model icon in the interactive controls of the control bar to switch between different model feedback information cards corresponding to different model icons in order to view the detailed feedback information of different model feedback information cards. In Design 2, users can click on the model icon in the interactive controls of the control bar to switch to the card containing the feedback information of the model corresponding to the clicked model icon. The interface after clicking is shown in the changing state diagram. In Design 3, users can hover over the control bar to bring up its interactive controls, as shown in the changing state diagram. In Design 4, users can click on the model icon in the interactive controls of the control bar to switch to the card containing the feedback information of the model corresponding to the clicked model icon. The interface after clicking is shown in the changing state diagram. 7. Other situations requiring explanation: Other explanation: "X" in the interface represents a text, number, or symbol area. Content represented by dashed lines in each design is not required to be protected.

The application discloses an interactive system and method based on digital audio and video family tree and family memory simulation, and aims to solve the problems of weak interaction of traditional family tree, rigid digital clone model and split memory personality. The system constructs an interactive family tree knowledge base containing member nodes, blood relationship and related data, and forms a structured family knowledge graph in multiple dimensions. The model graph co-evolution is proposed: multi-modal data is collected through daily interaction, incremental learning and knowledge graph updating are adopted in parallel, parameter efficient fine-tuning is used for lightweight learning of the digital clone, only a small amount of parameters are updated to overcome forgetting, and new event relationships are extracted to update the graph. Through bidirectional indexing, the personality and memory semantics are consistent, the multi-modal interaction is relationship-aware and dynamically evolved. The application synchronously updates the personality and memory, solves the problem that the digital clone cannot continuously evolve with the family, and realizes stereoscopic inheritance of family memory.

The invention discloses a power distribution network load flow calculation method based on graph synchronization parallel and forward-backward substitution, and the method comprises the steps: building a graph model in a graph database based on a power network, and enabling the graph model to be divided into a plurality of sub-graph models; setting an initial state for nodes in the sub-graph model, traversing the sub-graph model in a graph node parallel computing mode, and recording a node layer number and a node path of the sub-graph model; and in combination with the sub-graph model, completing graph calculation through forward tracking current calculation and backward tracking voltage calculation in sequence by utilizing a forward-backward substitution method, and judging whether an absolute value of a voltage difference value of nodes in the sub-graph model is smaller than the maximum iteration convergence precision or not. According to the method provided by the invention, the graph model is established, and graph calculation is carried out in a graph parallel mode in combination with a forward-backward substitution method, so that the calculation speed can be obviously improved.

The invention provides an optimization method and device for a multi-modal large model picture recognition cue word, and is applied to the technical field of data processing. The method comprises the following steps: firstly, extracting a preprocessed initial basic cue word and massive to-be-recognized pictures, and generating an initial cue word set and standardized picture data; identifying and screening a target picture by the model for the first time and generating scene matching json information, and separating a correct sample set from a misjudgment sample set through manual secondary verification; performing error analysis by utilizing the misjudgment sample set to generate a new cue word and difference degree data, and dynamically adjusting a scene cue word weight in combination with the number of samples; and the initial weight is newly established and configured for the scene with the too low weight, and the first 9 high-weight cues are screened for the next round of recognition, so that a cycle is formed to adapt to the spatial-temporal change, and the recognition accuracy is improved.

The application belongs to the technical field of computer vision and generative artificial intelligence, and particularly relates to a kind of holding digital model generation method and system based on conditional diffusion model, the relative position relationship of model joint and contrast article is mapped to the corresponding joint node in model graph, and the target position data of corresponding commodity is obtained;According to the target scaling factor, the goods to be displayed in the goods graph are scaled, and the goods scaling features are obtained;Using conditional diffusion model for conditional diffusion, the image of the goods held / worn by the digital model according to the posture reference graph is generated as the holding model image;Using the contrast article in the posture reference graph as the position and scale reference, precise scale control is realized;Ensure that the scale of the goods graph in the model graph is natural, solve the scale distortion problem in the traditional method;The posture reference graph is forced to include the contrast article, and the outline and key points thereof are used as position references, without manual adjustment, to improve the degree of automation and generation accuracy.

The invention discloses a graph chain thought-based large language model backdoor detection method, equipment and a medium. The method comprises the following steps of: constructing a graph chain type thinking template database; according to the current output response of the target large language model, retrieving N graph chain type thinking templates with the first similarity from a graph chain type thinking template database; constructing a model graph generation query according to the model graph generation cue word, the N graph chain type thinking templates and the current output response of the target large language model; respectively inputting the model graph generation query into a target large language model and a plurality of reference large language models to generate an inference graph; according to the inference graph discrimination prompt word, the inference graph generated by the target large language model and the inference graph generated by each reference large language model, constructing an inference graph discrimination query; and inputting the reasoning graph discrimination query into the judgment large language model, and analyzing the difference between the reasoning graph generated by the target large language model and the reasoning graph generated by each reference large language model to obtain a backdoor detection result.

The application relates to the technical field of BIM intelligent design and deep learning, in particular to a BIM model graph structure representation and model error identification method and device, wherein the method comprises the following steps: extracting semantic features, spatial features and topological features among various components according to component information of different types of components in a BIM model; generating corresponding graph structure representations based on the semantic features, the spatial features and the topological features, and constructing a training data set by using the graph structure representations; training a pre-trained graph neural network model by using the training data set to obtain a trained training model, and performing transfer learning by using the trained training model to identify an error identification target of the BIM model by using a transfer learning model. Thus, the problems that in the related art, the complex semantic, spatial and topological features of the BIM model bring difficult-to-ignore challenges, and the pre-training and transfer learning model cannot be perfectly suitable for the BIM model features are solved.

The application discloses a fault diagnosis method and system based on hard verification and an evolvable graph neural network, which obtains a final category cluster set through unique mode hard verification and mixed mode hard verification, and trains an expert model for each category cluster in the final category cluster set; then, a graph neural network is trained by using historical samples and the expert models; and then, a new category cluster is added by using real-time samples, the expert models and the graph neural network, and fault classification is performed. The application can autonomously and reliably discover new fault types under unsupervised conditions, seamlessly integrates the newly discovered fault types into a diagnosis model, realizes dynamic expansion of a model structure and self-evolution of a knowledge system, and has high robustness and adaptability to working condition changes.

This invention relates to a data tracking and management device for changes in land spatial planning and implementation, particularly in the field of land spatial planning technology. It includes: an image generation module for generating a plan view based on demarcated ecological protection red lines, permanent basic farmland, and urban development boundaries; a modeling module for modeling based on the generated plan view; a data acquisition module for acquiring the change in feature data of a single model map, the number of related features in two related model maps, and the spacing between the models; a data analysis module including interconnected first, second, third, and fourth computational units; and a parameter adjustment module for determining the comprehensive tracking frequency of several feature data based on a comparison of the comprehensive change parameters calculated by the first computational unit with the change parameter standard. This further improves the efficiency of land spatial planning.

An object of the system according to the embodiment is to efficiently organize information necessary for business planning and proposals and provide the information in a visually easy-to-understand format.SOLUTION: A system includes a reception unit, an analysis unit, a generation unit, and a provision unit. A reception part inputs information necessary for business planning and proposal. The analysis unit analyzes the information input by the reception unit, and lists characters and their relationships, how to hold and flow data, and unexamined points. The generation unit generates a business model diagram and a check list based on the information obtained by the analysis unit. The providing unit provides the output generated by the generating unit in a presentation format.SELECTED DRAWING: Figure 1

The present application relates to the technical field of BIM, and particularly relates to a BIM model drawing marking method, device, equipment and readable storage medium, the method comprises the following steps: obtaining all graphic element information in a BIM model, graphic element origin coordinates corresponding to each graphic element information and size parameters corresponding to each graphic element information; classifying and numbering each graphic element information according to a graphic element name to obtain a classification result; positioning the position of a component according to the graphic element origin coordinates corresponding to each graphic element information to obtain position information corresponding to each graphic element information; obtaining a drawing with completed size marking and graphic element attribute marking according to the size parameters corresponding to each graphic element information, the position information corresponding to each graphic element information and the classification result; the present application realizes the rapid and automatic marking of a BIM model drawing based on a BIM model to meet the requirements of engineering application, thereby making the design of a specified model standard box girder automatic and intelligent, and greatly improving the design efficiency of a high-speed railway standard box girder.

A computer-implemented method for modeling graph-structured data using a graph transformer includes applying cross-type attention to a target node of the graph-structured data. The graph transformer is connected to a type-specific feed forward network to allow node features of different node types to be learned differently. Masked label embedding is applied to force learning and a prediction of a node label for a node having a masked label.

A SoC multi-channel NPU performance evaluation method, device, equipment and medium, including dividing the model graph corresponding to the target neural network model to be run on the NPU to be evaluated to obtain a plurality of network model task segments; the execution process of each network model task segment is decomposed to obtain a plurality of execution stages with pipelining characteristics; the control rules of each execution stage are determined based on the preset attribute parameters of the network model task segment; the resource competition and concurrent scheduling rules between different execution stages are determined according to the key performance parameters of the NPU to be evaluated; the target neural network model is simulated and run based on the network model task segment and the execution stage through the above rules; the performance evaluation result of the NPU to be evaluated is determined based on the simulation running result, the key performance parameters and the SoC environment parameters carried by the NPU to be evaluated, so as to effectively improve the efficiency and performance fidelity of NPU performance evaluation under SoC multi-channel.

This application provides a distributed server rendering method, system, and medium. The method includes: component-based modeling and metadata definition of the rendered page using a component modeling module to obtain a component dependency model graph; acquiring user rendering request information; analyzing this information using a caching decision module to obtain a caching decision strategy; identifying independent rendering fragments using a fragmented rendering module based on the component dependency model graph; performing server-side fragmented rendering to obtain rendering fragment content packages; and performing multi-channel parallel transmission using a collaborative transmission module. Upon receiving the data, the client interaction module performs DOM rendering and data association binding to restore page interactivity. This application achieves distributed server rendering by constructing a component dependency model graph and defining metadata, making dynamic caching decisions based on real-time context and a preset component caching benefit prediction model, and combining fragmented rendering and multi-channel collaborative transmission with progressive data association and interaction recovery.

The application discloses a task-driven intelligent agent dynamic decision optimization method and system, which comprises the following steps: modeling graph structure data, constructing an adjacency matrix, a feature matrix and a similarity matrix; constructing neighborhood intelligent agents and layer number intelligent agents; based on the neighborhood intelligent agents, evaluating the importance of neighbor nodes through an attention mechanism and combining the upper confidence bound method to make reinforcement learning decisions, so as to obtain an optimal neighbor selection strategy; based on the layer number intelligent agents, dynamically determining the optimal aggregation network layer number for each node to obtain an optimal network layer number identification strategy; based on the above two strategies, re-normalizing the attention weight of the node and executing task-specific information aggregation under a graph neural network framework to generate a node embedding representation; constructing a joint loss function comprising a reconstruction loss and a self-supervised loss, optimizing the joint loss function, training the parameters of the neighborhood intelligent agents, the layer number intelligent agents and the graph neural network, and outputting an optimal node embedding result.

The application discloses an oil pipeline leakage monitoring method, device, equipment and medium, and relates to the technical field of pipeline monitoring. The method comprises the following steps: constructing a graph model of a pipeline monitoring network, wherein each node in the graph model is a measuring point on a pipeline; collecting pipeline data of each measuring point in real time, wherein the pipeline data comprises temperature, pressure and flow data; performing feature extraction on the pipeline data to obtain a plurality of feature data; determining a pipeline working condition for the plurality of feature data corresponding to each measuring point; adjusting the fusion weight of the fusion of various feature data in real time based on the pipeline working condition by using a reinforcement learning algorithm; performing feature fusion on the plurality of feature data according to the fusion weight to obtain multi-source fusion features; inputting the multi-source fusion features of each measuring point into a sensor fault recognition model to output a sensor fault label; and inputting the multi-source fusion features of each node carrying the sensor fault label into a leakage recognition and positioning model to output a leakage probability of each node. The application can improve the prediction accuracy.

The application provides an image style migration method and system based on WebGPU technology, which comprises the following steps: S1, constructing an image style migration model; S2, training the image style migration model; S3, exporting the trained and optimized image style migration model into an ONNX format, and then converting the image style migration model in the ONNX format into an intermediate format suitable for a Web-side inference engine, while generating necessary metadata and model atlas; S4, integrating a large language model, and automatically generating or optimizing style migration parameters and prompt words according to a style description or specified requirements input by a user; S5, integrating and deploying a complete image style migration model, and integrating and aligning the complete image style migration model with a prompt word optimization model to obtain a final image style migration model and convert the final image style migration model into a browser-side model suitable for WebGPU technology. The application realizes real-time migration of multi-style high-resolution images, and greatly improves the style migration speed and visual expressiveness.

The invention discloses a DFD and PLC code bidirectional tracing and dynamic mapping method and device based on IR, and relates to the technical field of electric data processing. The method comprises the following steps: acquiring a DFD primitive set file and a PLC code; respectively constructing an IR sub-graph model and an IR graph model; nodes and edges in the IR graph model are bound with source metadata; on the basis of the IR graph model, executing bidirectional semantic conversion of converting the DFD primitive set file into a PLC code in the forward direction and converting the PLC code into the DFD primitive set file in the reverse direction; consistency verification is carried out on the input and output behaviors before and after bidirectional conversion and the key path; constructing a ternary mapping index table for recording a corresponding relation among DFD elements, IR nodes and code positions; a monitoring agent is injected into a PLC runtime environment, runtime data is collected in real time, and the collected runtime data is bound to an IR graph model node and synchronously mapped to a DFD visual interface. According to the invention, the accuracy and reliability of full-life-cycle management of the control system are improved.

The application relates to the technical field of computer 3D modeling, and discloses a city three-dimensional terrain rapid modeling method and system based on AI point cloud semantic segmentation. According to city terrain semantic point cloud fusion preprocessing data set and in cooperation with a point cloud three-dimensional modeling platform, high-quality three-dimensional modeling of feature objects in a target city terrain is carried out, and precise modeling under a city terrain limited point cloud data state is realized based on semantic segmentation and point cloud fusion. Through the point cloud three-dimensional modeling platform and screenshot software, city terrain three-dimensional model image information is accurately collected, and intelligent identification of city terrain three-dimensional model modeling defects is carried out in combination with AI algorithms and city terrain feature object standard modeling defect image data. Based on city terrain three-dimensional model data and in cooperation with the point cloud three-dimensional modeling platform, dynamic and efficient repair and reconstruction processing of city terrain three-dimensional model modeling defects is carried out, and precise modeling and repair and reconstruction under a city terrain limited point cloud data state are realized based on model repair.