25 results about "Direct transformation" patented technology

The invention discloses a whole-body collaborative voice control system and method for a humanoid robot, relates to the technical field of robots, and aims to solve the problems of semantic understanding superficial layer, action planning localization and safety control lagging in existing humanoid robot voice control. Through closed-loop cooperation of the multi-mode intention understanding module, the whole-body action planning module, the dynamic balance security arbiter and the microkernel security scheduler, a unified control architecture is constructed. According to the system, natural language instructions are deeply analyzed into whole-body motion constraints through a semantic-action mapping knowledge base, anthropomorphic collaborative motion trails are generated, pre-security verification is conducted through a boundary model based on Lyapunov stability, and finally hard real-time execution and full-cycle security guarding are achieved through a microkernel security scheduler. According to the method, direct conversion from natural language to safe, stable and anthropomorphic whole-body motion is realized, and the task execution capability and the interaction naturalness of the humanoid robot in a complex scene are remarkably improved.

Systems and methods translate and govern directives across heterogeneous platforms. In some embodiments, directives are normalized into a common intermediate representation (CIR) for validation, conflict handling, and deployment. In other embodiments, a direct semantic mapping engine converts source directives directly into target-native, machine-executable directives without a CIR and without normalization, consulting a platform capability database and synthesizing target-compatible scripts, profiles, workflows, or rules where native primitives are absent. Ephemeral triggers (e.g., time window, incident, risk, posture, geofence) cause generation and atomic activation of temporary directives that supersede baseline directives for a defined scope and interval, with withdrawal on expiry to restore the baseline. Adapters actuate across families including cloud control planes, operating systems (Windows / macOS / Linux), IoT / OT, Kubernetes, identity, email / DLP, browser, and SaaS / social platforms. Both approaches share validation, simulation, and governance / deployment engines supporting approvals, audit logging, staged rollout, rollback, and continuous feedback.

Systems and methods for performing direct conversion of image sensor data to image analytics are provided. One such system for directly processing sensor image data includes a sensor configured to capture an image and generate corresponding image data in a raw Bayer format, and a convolution neural network (CNN) coupled to the sensor and configured to generate image analytics directly from the image data in the raw Bayer format. Systems and methods for training the CNN are provided, and may include a generative model that is configured to convert RGB images into estimated images in the raw Bayer format.

The application provides a kind of building door and window intelligent arrangement method and device, the method comprises: the geometric coordinates of building plan layout are mapped into discrete coordinate tokens, and the token sequence representing the geometric constraint information of building space is constructed according to the discrete coordinate tokens;The token sequence is input into the pre-trained multimodal large model to predict and generate the vector component sequence in the target area in a self-recurrent manner, wherein the multimodal large model establishes the semantic mapping relationship between the building space constraint and the component generation logic by learning the distribution probability and topological prior of building space layout, and the vector component sequence includes wall tokens and associated door and window tokens to represent the physical topological dependency relationship between wall and door and window components;The vector component sequence is restored to obtain the building vector design scheme, and the method realizes the direct conversion from discrete space representation to structured vector components, significantly improves the automation production efficiency and intelligent level of building layout design.

According to the automatic label conversion method, system and equipment and the storage medium provided by the invention, the method comprises the steps of scanning the first label to obtain the first label data, matching the supplier exclusive RF analysis rule, splitting and analyzing the attribute feature and the numerical feature, selecting the conversion mode according to the demand, associating the customer label rule, and performing automatic conversion on the customer label. Mapping sorting and integration of attribute values are completed to obtain second label data, a label template is customized according to requirements, data are filled to generate a graphic code, then the graphic code is integrated to generate a second label, and the second label is printed and archived, so that full-process automatic conversion from original label data to customer exclusive available labels is realized; according to the method, multi-scene requirements of single-label direct conversion, multi-label combination, single-label splitting and the like are met, the automation degree of label generation is improved, the error rate of data analysis and conversion is reduced, and the personalized requirements of multiple suppliers and multiple clients are met.

Systems and methods for performing direct conversion of image sensor data to image analytics are provided. One such system for directly processing sensor image data includes a sensor configured to capture an image and generate corresponding image data in a raw Bayer format, and a convolution neural network (CNN) coupled to the sensor and configured to generate image analytics directly from the image data in the raw Bayer format. Systems and methods for training the CNN are provided, and may include a generative model that is configured to convert RGB images into estimated images in the raw Bayer format.

The application discloses a humanoid robot full-body cooperative voice control system and method, relates to the technical field of robots, and aims at the problems of semantic understanding of the existing humanoid robot voice control being shallow, action planning being localized, and safety control being lagged. The application constructs a unified control architecture through the closed-loop cooperation of a multi-modal intention understanding module, a full-body action planning module, a dynamic balance safety arbitrator and a microkernel safety scheduler. The system deeply analyzes natural language instructions into full-body movement constraints through a semantic-action mapping knowledge base, generates a personified cooperative movement trajectory, carries out prior safety checking by using a boundary model based on Lyapunov stability, and finally realizes hard real-time execution and full-cycle safety guardianship through the microkernel safety scheduler. The application realizes direct conversion from natural language to safe, stable and personified full-body movement, and significantly improves the task execution capability and interactive naturalness of the humanoid robot in a complex scene.

The invention provides an off-line and on-line integrated feature derivation method and a related device. The off-line and on-line integrated feature derivation method is applied to a feature derivation system. The method comprises the following steps: determining a target derivation mode, wherein the target derivation mode comprises an offline mode and an online mode; determining a feature package and a derivation task according to the target derivation mode; the feature packet comprises one or more of original data, derivative data and custom data; performing derivation calculation on the feature packet according to the derivation task to generate a feature wide table; the feature wide table comprises a plurality of derivative features and is used for model training or data analysis; wherein the derivative calculation modes of the offline mode and the online mode are the same. Therefore, the same set of calculation scheme can be used for both off-line derivation and on-line derivation, so that off-line and on-line data can be directly converted, and the off-line and on-line data conversion efficiency is improved.

The application provides a GIS-based township land space planning base processing method and system, which comprises the following steps: extracting the past land class code from the three-adjustment plot vector to obtain direct conversion vectors and indirect conversion vectors, and updating the direct conversion vectors into first updated three-adjustment plot vectors; updating the indirect conversion vectors into second updated three-adjustment plot vectors and to-be-planned vectors through three types of data; updating the to-be-planned vectors into third updated three-adjustment plot vectors, and then obtaining final three-adjustment data. Compared with the manual processing method, the direct conversion vectors are converted into the first updated three-adjustment plot vectors, which avoids repeated calculation of most stable land blocks, and the coding of the indirect conversion vectors can be automatically assigned, which improves the base processing efficiency. Moreover, the three types of data with dynamic and real-time characteristics are introduced to make up for the tardiness of the three-adjustment data and improve the accuracy.

The application provides an ONNX neural network rear-end deployment conversion method and device. All operators in an ONNX model are traversed to find Matmul operators and Upsample operators, the Matmul operators are converted into corresponding dense operators or first combination operators or second combination operators in a relayir model, the Upsample operators are converted into corresponding resize operators in the relayir model, and the remaining operators are all directly converted into relayir operators, so that the ONNX model is converted into the relayir model, all operators can be supported by the rear end, the situation that operation cannot normally run is avoided, and the purpose of rapid deployment is achieved.

The present invention discloses a virtual-real combined TSN network flow scheduling simulation method and related equipment, the method comprising: obtaining device model and device capability parameters, adding virtual devices based on the device model, and instantiating the virtual devices based on the device capability parameters to obtain instantiated devices; creating a topology template library, selecting a target topology template based on the topology template library, importing the instantiated device into the target topology template, and configuring flow information for the target topology template to select a scheduling strategy; calculating the scheduling strategy based on the TSN scheduling algorithm, and streaming the calculation results to the devices of the actual network to complete the operation of the actual TSN network services in the devices of the actual network. The present invention can simulate the behavior and properties of real devices, and can also simulate or import real topology templates, and has no restrictions on the algorithm language, and the output of the network scheduling algorithm can be directly converted into real device configuration.

The invention provides a small sample projection coordinate direct conversion method based on neural network classification judgment and related equipment. The method comprises the following steps: acquiring projection coordinates under a first reference system to be converted; for the projection coordinates to be converted under the first reference system, obtaining the projection coordinates of the projection coordinates under the second reference system through polynomial fitting; inputting the projection coordinates in the first reference system and the projection coordinates, obtained by fitting, in the second reference system into a trained neural network model to judge whether the distance conversion error is within a preset distance error range, and outputting the distribution condition of the conversion errors in the horizontal and vertical directions and corresponding probability values; and if the distance conversion error is not in the preset distance error range, correcting the output probability value based on the Bayesian law and Monte Carlo simulation, and correcting the projection coordinate under the second reference system obtained by fitting based on the corrected probability value. According to the invention, the precision of single projection coordinate conversion in the statistical sense is improved.

Systems and methods translate and govern directives across heterogeneous platforms. In some embodiments, directives are normalized into a common intermediate representation (OR) for validation, conflict handling, and deployment. In other embodiments, a direct semantic mapping engine converts source directives directly into target-native, machine-executable directives without a OR and without normalization, consulting a platform capability database and synthesizing target-compatible scripts, profiles, workflows, or rules where native primitives are absent. Ephemeral triggers (e.g., time window, incident, risk, posture, geofence) cause generation and atomic activation of temporary directives that supersede baseline directives for a defined scope and interval, with withdrawal on expiry to restore the baseline. Adapters actuate across families including cloud control planes, operating systems (Windows / macOS / Linux), loT / OT, Kubernetes, identity, email / DLP, browser, and SaaS / social platforms. Both approaches share validation, simulation, and governance / deployment engines supporting approvals, audit logging, staged rollout, rollback, and continuous feedback.

The present invention relates to a method for rapid prediction of urban waterlogging distribution based on an improved convolutional neural network, comprising: 1. constructing a rapid prediction model data set: the data set is a waterlogging influencing factor-waterlogging characterization factor data set, and the waterlogging influencing factors include rainfall factors, terrain factors, pipe network length factors, and distance factors from the river system; 2. building a rapid prediction model based on a convolutional neural network and performing training and testing. In the process of building the model, different forms of waterlogging influencing factors are first encoded separately, then fused, and finally decoded after fusion; in the process of training and testing, an improved loss function is used; 3. model prediction and model prediction result analysis. This method accurately realizes the direct conversion of rainfall and flood elements (flooding distribution, maximum flooding depth), and improves the speed and accuracy of urban flood forecasting.

The invention relates to an end-to-end cross-CPU (central processing unit) architecture application program adaptation method. The method comprises the following steps: acquiring data; carrying out loading and feature extraction on the data; meanwhile, feature embedding is carried out according to the extracted features, and a BinCode-Transform model is constructed and trained; carrying out the semantic reduction of the embedded vector through a trained BinCode-Transformer model, and carrying out the semantic reduction of the embedded vector through the trained BinCode-Transformer model; and generating a file according to an embedded vector semantic reduction result. Due to the fact that semantic expression space of structures such as ELF or PE of binary files is small, the basic forms are operation codes, operands and keywords of file structures, the model trained through the method is small (the space is much smaller than that of natural language complexity), the model can run on a CPU, and cost is saved. Meanwhile, binary files for different CPUs are directly converted through artificial intelligence, operation is easy, the cost and threshold of a bottom layer library and application multi-CPU adaptation are reduced, the efficiency of localization adaptation of different applications is greatly improved, and ecological chain development of localization credential products is promoted.

The invention provides a GIS-based village and town land space planning cardinal number processing method and system, and the method comprises the steps: extracting previous land class codes from a triple-tone pattern spot vector, so as to obtain a direct conversion vector and an indirect conversion vector, and updating the direct conversion vector into a first updated triple-tone pattern spot vector; updating the indirect conversion vector into a second updated triple-tone pattern spot vector and a to-be-planned vector through three types of data; and updating the to-be-planned vector to a third updated triple-tone pattern spot vector to obtain final triple-tone data. Compared with a manual processing mode, the direct conversion vector is converted into the first updated triple-tone pattern spot vector, repeated calculation of most stable plots is avoided, code endowing of the indirect conversion vector can be automatically and completely completed, cardinal number processing efficiency is improved, three types of data with dynamic performance and real-time performance are introduced, and the processing efficiency is improved. The hysteresis of three-tone data is compensated, and the accuracy is improved.

The invention discloses a novel agrobacterium carrier system and application thereof. The invention provides an agrobacterium vector system, and the agrobacterium vector system comprises an expression vector and an auxiliary vector, the expression vector comprises elements such as a pVS1 oriV nucleotide sequence of a replication start site from a pVS1 replication system and the like; the auxiliary vector comprises a nucleotide sequence from a replication start site oriV of a pRK2 replication system and other elements. The agrobacterium vector system provided by the invention can keep high copy in escherichia coli and agrobacterium respectively, has very high plant transformation efficiency and very small skeleton, is very easy and flexible in construction, and is very suitable for construction of libraries. Moreover, as the framework is small and the construction efficiency is high, the step of transforming escherichia coli can be omitted, the agrobacterium can be directly transformed, and the positive rate is high, so that the whole experimental period is shortened, and the workload is reduced.

The invention provides a courseware conversion method and device, equipment and a storage medium, and the method comprises the steps: obtaining a to-be-converted two-dimensional courseware, and determining a target three-dimensional scene corresponding to the two-dimensional courseware; and converting a courseware page of the two-dimensional courseware into a three-dimensional area of the target three-dimensional scene to obtain a three-dimensional courseware corresponding to the two-dimensional courseware. According to the technical scheme, direct conversion from the two-dimensional courseware to the three-dimensional courseware can be realized, the three-dimensional courseware can be obtained without using special three-dimensional making software by a professional maker, the required time is short, and the courseware conversion efficiency can be improved.

The invention relates to a coordinate system conversion method based on digital twinning, and the method comprises the steps: planning a coordinate conversion path through a conversion scheduling module according to the combination of a source coordinate system type and a target coordinate system type; a corresponding conversion unit is called through a conversion algorithm module to execute conversion, and the conversion unit comprises a WGS84 coordinate system and GCJ02 coordinate system conversion unit, a GCJ02 coordinate system and BD09 coordinate system conversion unit and an approximate reverse calculation unit. And if the target is the UE engine coordinate system, a result is output through the output adaptation module, if the target is the UE engine coordinate system, the WGS84 coordinate system coordinates are transmitted to the Cesium docking module, and conversion with the UE engine coordinate system coordinates is completed through a CesimGeoreence assembly. Flexible conversion among a WGS84 coordinate system, a GCJ02 coordinate system, a BD09 coordinate system and a UE engine coordinate system is supported, and the industrial defect that the BD09 coordinate system cannot be directly converted into the WGS84 coordinate system is particularly overcome. And meanwhile, the method adapts to UE4 and UE5 series engines, and can meet diversified requirements of POI point position marking, track route planning and the like in a digital twin scene.

The application provides a production process planning and programming method, device and system based on material flow, which comprises the following steps: S1, describing a flow of a production process, wherein the flow comprises a plurality of process steps with sequence; S2, setting a material state information between each process step and at the beginning and end of the whole flow respectively; S3, generating a material state description graph by using a plurality of ontology elements, wherein the material state description graph comprises material, material property and material relationship. The application fills the gap of production process planning and equipment end programming, so that the material flow state description in the production process planning can be directly converted into input parameters of the equipment end programming, and the whole process is more efficient and convenient.

The invention provides a tensor conversion path automatic planning method for a heterogeneous chip, belongs to the technical field of computer software and artificial intelligence chips, and constructs a three-level routing strategy including'direct conversion ', 'equipment end rearrangement' and'host rollback '. A compiler automatically judges whether the hardware can meet a conversion request or not through a preset hardware capability graph; the dependency deadlock is avoided preferentially by adjusting an atomic operation sequence (for example, changing layout first and then converting type) at a device end; and when the equipment end path is completely unreachable, an equipment-host-equipment rollback path is automatically generated, and the universality of the host is used as a transfer springboard. According to the mechanism, the optimal instruction sequence considering the execution efficiency and the program correctness is automatically generated under the condition that manual intervention is not needed.

The invention discloses a polarization rotator and a design and preparation method thereof. The polarization rotator comprises a straight waveguide structure, a groove is formed in the top of a waveguide, the width of the bottom edge of the waveguide is gradually widened / narrowed from the starting end to the tail end, and the eccentric distance between the middle point of the bottom edge of the waveguide and the middle point of the bottom of the groove is gradually widened from 0 to the maximum value and then gradually narrowed to 0. According to the structural design, direct conversion between two different polarization basic modes TM0-TE0 can be realized in the straight waveguide without a bending structure, the conversion efficiency is high, and the performance is excellent; the device is simple in structure and small in transverse size; according to the invention, the waveguide structure with non-uniform thickness can be formed only through one round of photoetching and etching process, so that the processing flow is greatly simplified, the processing cost is reduced, and the processing error caused by overlay deviation limitation among multiple rounds of photoetching and the influence on the performance of the polarization rotator are avoided; meanwhile, the polarization rotator is close to the strip-shaped waveguide in structure, and energy loss in the transition process of the polarization rotator and the strip-shaped waveguide is small.

The invention discloses a data format conversion method and device, equipment, a storage medium and a program product, and relates to the technical field of data processing. The method comprises the steps of obtaining source data to be subjected to data format conversion, a source data format corresponding to the source data and a target data format; and based on the source data, the source data format and the target data format, determining to directly convert the source data from the source data format to the target data format by adopting a target conversion mode. According to the method, the source data format can be directly converted into the target data format from the source data format only through single-step conversion, intermediate resources do not need to cache intermediate results, logic implementation is simple, and universality is high.

The invention relates to a method for realizing credential and credential adaptation based on a decompilation Decom-Transform model. The method comprises the following steps: acquiring data; carrying out loading and feature extraction; carrying out feature embedding on the extracted features, and constructing a Decom-Transform model to obtain an Encoder model and a Decoder model; an Encoder model and a Decoder model are used for generating an assembly language; generating a large language model according to the generated assembly language; and finally, generating a high-level language by using the generated large language model. Due to the fact that semantic expression space of structures such as ELF or PE of the binary file is small, and the basic forms are operation codes, operands and keywords of file structures, the trained model is small and can run on a CPU, and cost is saved. Meanwhile, the decompiling efficiency is high, and the speed is high; the binary file is decompiled into the wanted high-level language through artificial intelligence direct conversion, operation is easy, learning of the structure and development logic of underlying software is facilitated, and ecological chain development of domestic credential products is promoted.

This invention discloses a document information extraction method, system, and program product that combines logical constraints and reinforcement learning, relating to the fields of artificial intelligence and intelligent document processing technology. Firstly, by explicitly incorporating arithmetic rules into the reward function, this invention significantly reduces "numerical illusions" and substantially improves logical consistency. Secondly, it solves the problem of extracting borderless tables, deriving the table structure through semantic content and logical constraints instead of relying on visual dividing lines, making it particularly suitable for borderless complex tables such as medical invoices. Thirdly, this invention uses reinforcement learning to transform discrete business rules into dense supervisory signals to guide the optimization of the document information extraction model. Furthermore, the document information extraction model of this invention directly outputs JSON format conforming to the database schema, eliminating the need for complex post-processing steps and achieving direct conversion from images to a queryable database.