66 results about "Visual query" patented technology

In an example, a method for fine-tuning a Large Visual Language Model (LVLM) includes providing visual queries, each of the visual queries comprises at least an image and a textual query related to the image; processing, by the LVLM, the visual queries to extract visual embeddings from the visual queries, wherein the LVLM comprises a Visual Language Model (VLM), a first Large Language Model (LLM), and a linear projection layer interconnecting the VLM and the LLM; for visual queries: i) generating, by the LVLM, a response to the corresponding visual query based on the corresponding visual embedding; ii) evaluating, by a second LLM, the generated response to verify that the generated response satisfies predefined criteria; and iii) providing, by the second LLM, a feedback to the LVLM, in response to the evaluating the generated response; and fine-tuning the LVLM using aggregated feedback provided by the second LLM for the visual queries.

The invention belongs to the technical field of mine ventilation monitoring, and aims to solve the problems that a traditional knowledge system based on a rule base has bottlenecks in the aspects of dealing with sudden working conditions, knowledge updating and semantic reasoning, and is high in construction cost and complex to maintain. The invention provides a mine ventilation knowledge graph construction method based on a large model, and the method comprises the following steps: S100, obtaining and cleaning a multi-source heterogeneous text related to mine ventilation, and obtaining a JSON format knowledge fragment of a unified structure; s200, constructing an ontology model and defining a semantic structure; s300, constructing four types of knowledge extraction tasks, and extracting entities, attributes and relationships; s400, constructing an entity alignment module; and S500, constructing a graph database structure. According to the method, automatic structured expression, semantic consistent fusion and intelligent visual query of knowledge in the mine ventilation field can be realized, and an interpretable, extensible and reasonable intelligent support platform is provided for a mine ventilation system.

The invention relates to a Chinese language and literature computer networking query reading system. According to the system, a query request processing module carries out mixed feature analysis processing according to a query request of a user, and a mixed query vector fusing radicals and semantics is generated; a distributed retrieval module performs distributed retrieval in a preset space-time fragment index structure based on the mixed query vector to obtain a candidate literature data set; the preliminary retrieval optimization module is used for carrying out self-adaptive conversion processing on abnormal characters and traditional and simple characters in the candidate literature data set according to a Chinese character evolution rule and a context to obtain standardized text data; and the visual query output module performs knowledge graph enhancement processing on the standardized text data to generate a visual query result. According to the system, efficient query and visual display of Chinese language and literature works are achieved through multi-modal processing, multi-dimensional indexing and knowledge graph technologies, and the problems that a traditional query mode is single in input, limited in retrieval dimension, insufficient in text processing and the like are effectively solved.

The invention discloses a traditional Chinese medicine decoction piece production data management method and system based on quality tracing, and relates to the technical field of production data management. According to the invention, a unique tracing identification code is generated for warehousing traditional Chinese medicine raw materials, and information is bound to create a digital file; in the production process, data and quality data of all procedures are collected in real time by scanning the identification codes, and a dynamic production digital file is formed; an improved TCN neural network is trained to construct a quality prediction model, and the model can accurately predict product quality by introducing optimization such as a process perception attention mechanism and parallel branches; performing real-time quality evaluation and process adjustment by using the model in production; key process data is stored through a block chain, and a traceability code is provided on a finished product package for visual query; data communication and credible tracing of a whole production chain are achieved, quality control is converted into active prediction and intervention from passive inspection, and the quality stability and management efficiency of the traditional Chinese medicine decoction pieces are remarkably improved.

The invention discloses a telemetering site position display and information query system and method based on AR information labeling. The system comprises a multi-source data management module, an interactive query module, an AR recognition module and a mobile terminal device. The interactive query module is arranged in mobile terminal equipment, and the mobile terminal equipment is in communication connection with the AR identification module through the communication module; the AR identification module is in communication connection with the multi-source data management module through the communication module; equipment pose synchronization, spatial range screening and a three-dimensional space anchoring algorithm based on ARCore are executed through the system, and finally hierarchical information capable of being interactively displayed is generated through AR label generation and rendering; the method aims at solving the problems that in the prior art, a telemetering site information query mode is backward, the information display dimension is single, and operation steps are tedious and low in efficiency, and telemetering site visual query can be achieved through the combination of the mobile intelligent device and the space positioning and AR labeling technology.

The invention provides a method for querying geometric elements in different geometric modeling engines and a related product, and the method comprises the steps: responding to a first user operation, and determining a target positioning UV parameter point coordinate in a UV parameterization plane; determining target geometric elements from the entity model according to the coordinates of the target positioning UV parameter points; and determining a target geometric modeling engine of the currently loaded entity model, and determining target identification information of the target geometric element in the target geometric modeling engine. According to the geometric modeling method and device, a user can select the points near the geometric elements needing to be operated by operating the interactive interface and map the points to the corresponding geometric elements through the coordinates corresponding to the points, so that the situation that the geometric modeling efficiency is greatly improved due to different identification modes in different geometric modeling engines can be avoided. And the problems of narrow use case application range, non-visual query result and the like existing in querying the geometric elements by using the ID are solved.

The invention provides a BIM shell extraction modeling method and system, and the method comprises the steps: analyzing a BIM in an IFC format into a structured fine model, reserving geometric, semantic and material information, carrying out the coding and simplification of each building component, and generating a simplified query model for shell extraction. Creating two cameras with consistent parameters in a three-dimensional scene, wherein the main camera is used for rendering a fine model; the auxiliary camera renders the simplified model in the background. And through the view transformation matrix, the viewpoint positions of the main camera and the auxiliary camera are kept synchronous. The auxiliary camera retrieves visible external building components through a rendering-to-texture technology, and remaps query results into the viewing angle of the main camera, so that the components are highlighted in a visual environment. According to the method, BIM shell components are selected in a real-time interactive visualization mode, the accuracy and efficiency of complex BIM shell extraction are improved, and visual query viewpoints do not need to be additionally arranged.

The invention discloses an NL2SQL and NL2VIS method based on large model collaborative intelligent agent, and belongs to the technical field of natural language processing. The NL2VIS method comprises the steps that candidate database modes related to natural language query are obtained from database modes, and task complexity corresponding to the natural language query is obtained according to the number of tables contained in the candidate database modes; in combination with the task complexity, natural language query is converted into a preliminary SQL query statement; verifying, repairing and feeding back the SQL query statement based on the initial SQL query statement until a final SQL query statement is generated; and converting the final SQL query statement into a visual query language. According to the method and the device, the database table, the field and the association relation of the database table and the field involved in the query intention of the user can be more accurately identified, and connection errors, field mistaken selection and information omission caused by mode misunderstanding are fundamentally reduced.

The invention relates to the technical field of data processing, and discloses a statistical report generation method and system, electronic equipment and a storage medium, and the method comprises the steps of counting a configuration request, formulating a statistical statement template, and creating a statistical task; triggering a batch processing process according to timing configuration, executing a statistical task, performing batch calculation on the source data set associated with the statistical statement template, generating a statistical result, and persistently storing the statistical result in a specified database; selecting a query mode according to the visual query request, an identifier corresponding to the statistical task and a time range selected by a user, and performing formatting and merging calculation on a query result to generate a visual data set; performing rendering calculation on the visual data set according to the visual template configuration, generating chart data and displaying the chart data; and identifying and migrating the historical statistical data, receiving a query request for the archived data through an independent query interface, and returning a corresponding result. According to the invention, diversified statistical requirements can be met, and the integrity and flexibility of data tracing are guaranteed.

The invention discloses an SAM2 small sample segmentation method based on semantic-visual dual-memory fusion, and the method comprises the steps: constructing semantic query memory, visual query memory and query-related support visual memory, and fusing the semantic query memory and the visual query memory through a memory refinement module guided by the query-related support visual memory. SAM2 dense matching and decoding module end-to-end training are combined. The problems of single memory and foreground-background confusion of an existing method are solved, target semantic consistency and fine-grained modeling ability are enhanced, segmentation precision and generalization ability are remarkably improved in a complex scene, and the method is suitable for the fields of medical image analysis, automatic driving and the like.

The application provides a kind of operation and maintenance scene visualization report generation method and system based on natural language question and answer, belongs to intelligent operation and maintenance technical field, method includes: receiving the natural language query input by user;Through the pre-training of large language model and the preset operation and maintenance terminology dictionary, the natural language query is parsed and entity is extracted, and the entity-field association table containing demand type is generated;According to demand type and semantic feature, the query type is judged to be data query or visual query;Based on the judgment result, generate structured query language sentence or visual instruction;Query is executed to the interface business database to obtain raw data;Raw data is processed and analyzed to obtain analysis result data;Call data feature adaptation algorithm to match chart type, generate and output visual report.The application realizes the full-link automation from natural language input to visual report output, reduces the operation and maintenance data interaction threshold, improves the operation and maintenance data processing efficiency and accuracy.

The invention relates to the technical field of detection and tracing based on the Internet of Things, in particular to a detection and tracing system based on the Internet of Things, which comprises a sensing layer, a network layer, a processing layer and an application layer. The sensing layer is integrated with a high-precision sensor to collect multi-dimensional data; the network layer transmits data through 5G or NB-IoT encryption; the processing layer stores and analyzes data by using a block chain and an artificial intelligence algorithm; and the application layer provides visual query and interaction functions. The system can achieve the real-time collection, reliable storage and intelligent analysis of full-chain information, overcomes the defects of an existing system in the aspects of information comprehensiveness, data safety and intelligent level, and meets the efficient, transparent and credible tracing requirements of the modern industry.

The invention relates to an infectious tophus ulcer wound nursing scheme assessment method and system, and the method comprises the steps: collecting the standardized electronic medical record data of a patient through a medical information system, including demostatistics, wound characteristics, laboratory inspection index sequences and nursing records; after structured processing, timestamp unification and missing value filling are conducted on the data, the data are input into a pre-trained deep learning model, expected wound healing probabilities, predicted healing time and complication risk scores corresponding to different nursing schemes are output, then a comprehensive assessment report is generated, the report is stored in a database, and a visual query interface is provided. And meanwhile, collecting actual nursing result data for continuous optimization and updating of the model. According to the invention, quantitative evaluation and prediction of the nursing scheme are realized, clinical decision making is effectively assisted, and the accuracy of nursing management and the resource utilization efficiency are improved.

The invention discloses a wireless communication resource management system and method, and the system comprises an equipment management and data processing module which is used for interacting with physical equipment, configuring the composition relation between virtual resources and the physical equipment through a database, and providing a data conversion processing interface between the virtual resources and the physical equipment; the virtual resource management module is used for storing and managing parameters and state information of virtual receiving resources and virtual transmitting resources; the resource allocation management module is used for allocation and recovery management of virtual resources and externally providing a resource application interface; the resource use management module is used for monitoring and managing the use condition of the virtual resources and providing a visual query interface; according to the method, the physical device is abstracted into the virtual resource, decoupling of the upper-layer application and the physical device is achieved, and the flexibility and expansibility of the system are improved.

The invention provides a density map data visual query method and device based on a flow model. The method comprises the following steps: aggregating spatio-temporal data related to a target area to obtain different data frames; based on a configured kernel density estimation algorithm, processing the different data frames to generate density maps corresponding to the different data frames and corresponding density map data; inputting the density map data into a trained flow model, and encoding the density map data corresponding to any data frame by the flow model to obtain an encoding result representing the density map data to the corresponding representation space; after the space-time range and the coding result involved by the data frame are stored in an associated mode, according to the query condition input by the user, the trained flow model is used for conducting visual query on the stored data set, and the query result is displayed to the user. The method can significantly reduce the space occupancy and can accelerate the efficiency of common aggregation operations.

A system may include a content repository that stores multi-modal content comprising text, visual content, and / or audio content. The system may include a processor programmed to: access a prompt comprising a text query, a visual query comprising text that describes a visual to be found, and / or an audio query comprising text that describes audio to be found, execute a language model based on the prompt to identify content from the content repository, receive, from the language model, a request for a callback function that seeks additional information to satisfy the multi-modal query, execute the callback function to obtain the additional information and provide the additional information to the language model in response to the request for the callback function, re-execute the language model based on the multi-modal query and the additional information, obtain, from the language model, content responsive to the prompt based on the additional information.

The embodiment of the invention provides a cultivated land quality monitoring-oriented data blood relationship construction and reasoning method and system, and is applied to the technical field of cultivated land quality monitoring and geographic information data management. The method comprises the following steps: acquiring multi-source heterogeneous cultivated land data of a target area, and organizing and standardizing the multi-source heterogeneous cultivated land data to generate standard cultivated land quality monitoring index data; according to the standard cultivated land quality monitoring index data, constructing a cultivated land quality evaluation system, and generating evaluation result data and consanguinity metadata; according to the evaluation result data and the blood relationship metadata, constructing a cultivated land quality blood relationship knowledge graph; and carrying out visualization processing on the cultivated land quality consanguinity knowledge graph so as to realize visual tracing and reasoning on the cultivated land quality monitoring data consanguinity. In this way, a full-link traceable blood relationship network can be constructed, and visual query and bidirectional reasoning can be realized based on the blood relationship network, so that full-link transparency and dynamic influence calculability in the evaluation process can be realized.

The invention particularly relates to a mobile data monitoring service system and method. The mobile data monitoring service system comprises a self-service terminal machine, a data statistics module and an operation monitoring module, the data statistics module is responsible for reporting and counting operation data and business handling data of the self-service terminal machine; and the operation monitoring module is responsible for performing unified monitoring management on self-service terminal machines distributed in various regions according to the regions, monitoring the operation state of the current machine in real time, accurately positioning an abnormal terminal when an abnormal operation condition is encountered, and then informing a manager and a maintenance task in time to perform management and maintenance processing. According to the mobile data monitoring service system and method, visual query of statistical data is realized, the operation states of all deployed terminal machines can be counted, macroscopic regulation and control can be carried out, maintenance personnel can be notified to deal with abnormal problems in time when the abnormal problems occur, the operation risk is reduced, and business propaganda and popularization can be carried out simply and conveniently.

Systems and methods of obtaining data from a cellular network are disclosed. In some embodiments, visual query options are presented in a graphical user interface (GUI), wherein the visual query options define a query. Query option selections of the visual query options are received. A query template is generated in accordance with the query option selections. An audit of parameters is performed related with network elements of the cellular network in accordance with the query template. A report of the audit is transmitted to a user device. A visual representation of the report is presented in the GUI.

The invention discloses a development method and system of a self-adaptive registration framework. The method comprises the steps that an original image is input into a registration model and a pre-trained three-dimensional medical visual basic model VFM, the registration model is used for predicting a deformation field, and the VFM is used for extracting feature vectors and generating segmentation masks by using a double-image visual query fusion BIVQF technology; fusing the output of the encoder and the feature vector by using a hierarchical attention fusion mechanism; according to the hierarchical attention fusion mechanism, W-MCA is used in a double-flow registration network, G-CAT is used in a single-flow registration network, and features are fused. Inputting a fused result into a registration decoder for image registration; and further training the registration network by using the segmentation mask, and updating parameters of a prediction head. According to the invention, the registration efficiency can be improved.

The present application relates to a kind of petrochemical emergency rescue standard knowledge graph construction and retrieval method, including collecting petrochemical product emergency rescue standard text, and the standard text is preprocessed;Design petrochemical product emergency rescue standard ontology, according to the standard knowledge of petrochemical product related standard design petrochemical product accident rescue standard entity, petrochemical product accident rescue standard relationship and knowledge graph bottom data structure;Standard rules are used to extract entity and relationship in the standard common standard elements in entity;Based on deep learning model, the core elements of standard text are entity-recognized and relationship-extracted;The obtained petrochemical product accident rescue standard entity, triple is fused into knowledge, and is stored in graph database;According to graph database, visual query and intelligent retrieval are carried out, the present application provides a kind of petrochemical emergency rescue standard knowledge graph construction and retrieval method, integrates the standard knowledge involved in rescue and provides high-quality knowledge, provides decision information for experts and command personnel.

A method is provided for interacting with audience members in an event, each of the potential attendees having available thereto a unique identifier. The method comprises creating, for an attendee, a unique ID (UID) on a mobile wireless device (MWD) by the steps of inputting to the MWD one of the unique identifiers, combining the obtained unique identifier with a UID time stamp at the time of creation of the UID; receiving with a server on a first wireless channel communications from the MWD; registering the UID at the physical location of the event; generating a visual query; displaying on the MWD response indicators; receiving at the server from the registered attendee a response, to the query over the first wireless channel; and storing in a database on the server the received response in association with the displayed query.

The application provides a density map data visual query method and device based on a flow model. The method aggregates spatio-temporal data related to a target region to obtain different data frames; based on a configured kernel density estimation algorithm, the different data frames are processed to generate density maps corresponding to the different data frames and corresponding density map data; the density map data is input into a trained flow model, the flow model encodes the density map data corresponding to any data frame to obtain an encoding result representing the density map data to a corresponding representation space; after the spatio-temporal range related to the data frame and the encoding result are associated and stored, a trained flow model is used to perform visual query on the stored data set according to a query condition input by a user, and the query result is displayed to the user. The method can significantly reduce the space occupation and can accelerate the efficiency of common aggregation operations.

The invention discloses a key process digital management and control system and method.The system comprises a process planning system, a product data management system and a manufacturing execution system, the process planning system completes compiling of a structured process file and carries out key process identification, and the structured process file comprises structured process data; the process data are recorded in a key process operation content table, a key process quality control table and a key process inspection record table; the product data management system archives the structured process data in an XML text form and issues the structured process data to the manufacturing execution system; and the manufacturing execution system receives and analyzes the process data in the XML text form, creates a structured three-fixed table, fills a production process operation result and a detection result in a key process inspection record table, and generates a visual record table. According to the invention, structured storage, electronic recording and visual query of key process data can be realized, so that digital management and control of the key process production process can be realized.

A system may include a content repository that stores multi-modal content comprising text, visual content, and / or audio content. The system may access a prompt comprising a text query that describes text to be found, a visual query comprising text that describes a visual to be found, and / or an audio query comprising text that describes audio to be found, execute a language model based on the prompt to identify content from the content repository, receive, from the language model, a request for a callback function that seeks additional information to satisfy the multi-modal query, execute the callback function to obtain the additional information and provide the additional information to the language model in response to the request for the callback function, re-execute the language model based on the multi-modal query and the additional information, and obtain, from the language model, content responsive to the prompt based on the additional information.

In an example, a method for fine-tuning a Large Visual Language Model (LVLM) includes providing visual queries, each of the visual queries comprises at least an image and a textual query related to the image; processing, by the LVLM, the visual queries to extract visual embeddings from the visual queries, wherein the LVLM comprises a Visual Language Model (VLM), a first Large Language Model (LLM), and a linear projection layer interconnecting the VLM and the LLM; for visual queries: i) generating, by the LVLM, a response to the corresponding visual query based on the corresponding visual embedding; ii) evaluating, by a second LLM, the generated response to verify that the generated response satisfies predefined criteria; and iii) providing, by the second LLM, a feedback to the LVLM, in response to the evaluating the generated response; and fine-tuning the LVLM using aggregated feedback provided by the second LLM for the visual queries.

The present application provides a garden plant geographical information query system and method based on information recognition, which relates to the field of smart garden applications. First, it extracts features from the image data of garden plants to obtain the morphological attribute features of garden plants. At the same time, it extracts features from the name and distribution location information of garden plants to obtain the distribution spatial geographical features of garden plants. Then, it fuses the morphological attribute features of garden plants and the distribution spatial geographical features of garden plants to obtain the garden plant features. Next, it optimizes the fused features. Finally, it uses a generator to generate a plant landscape spatial geographical query file from the optimized garden plant features. In this way, it can meet diversified query requirements and can also combine the generated plant landscape spatial geographical query file with a geographical information system to achieve visual query.

A user can submit a visual query that includes one or more images with user free-form selected visual features of interest. Various processing techniques such as optical character recognition (OCR) techniques can be used to recognize text (e.g., in the image, surrounding image(s), etc.) and / or various object detection techniques (e.g., machine-learned object detection models, etc.) may be used to detect objects and particular visual features of objects (e.g., dress, sleeves, color, pattern, etc.) within or related to the visual query. Content related to the detected text or object(s) in combination with the user free-form selected visual feature of interest can be identified and potentially provided to a user as search results. As such, aspects of the present disclosure enable the visual search system to more intelligently process a visual query to provide improved search results and content feeds, including search results which are personalized to account for user search intent.

The embodiments described herein are directed to computer-implemented visual query matching methods and systems for implementing thereof. An example method includes storing first query data including queries associated with a first set of users, wherein each query includes selected criteria from multiple query criteria and first filter data for the multiple query criteria; storing second query data including metadata and queries associated with a second set of users, wherein each query includes selected

The embodiments described herein relate to a computer-implemented visual query matching method and implementation system thereof. The example method includes storing first query data, the first query data including queries associated with a first user set, wherein each query includes a selected condition from a plurality of query conditions and first filtering data for the plurality of query conditions; storing second query data, the second query data including metadata and queries associated with a second user set, wherein each query includes a selected condition from a plurality of query conditions and second filtering data for the plurality of query conditions; determining, by a processor, matching users of a first user in the first user set, the matching users being from the second user set; and providing, by the processor, a dashboard display including a matrix of the matching users to the first user.