249 results about "Data Annotation" patented technology

The invention provides an intelligent data labeling method and system based on multi-modal fusion and large model verification, belongs to the field of artificial intelligence and data processing, and innovatively fuses multi-modal information such as an OCR recognition result, a layout structure, original image visual features and deep semantic analysis of a large language model (LLM). And a precise automatic labeling result credibility evaluation mechanism is constructed. According to the method, various errors in automatic labeling can be accurately recognized and adaptively corrected, and the errors comprise conventional error correction based on hard coding rules and complex semantic error correction driven by LLM. Meanwhile, the system can continuously optimize the data labeling capability of the system through an efficient man-machine cooperation and closed-loop feedback learning mechanism, and automatically precipitate domain knowledge assets. The invention aims to solve the problems of recognition accuracy bottleneck, heavy manual proofreading burden, lack of intelligent judgment and error correction, knowledge accumulation lag and the like in traditional document data labeling, so that the efficiency, accuracy and automation level of document data labeling are remarkably improved.

The invention relates to the technical field of computer vision and automatic driving, in particular to a self-supervised aerial view perception method fusing Gaussian spattering and time sequence modeling, electronic equipment and a readable storage medium, and the method comprises the following steps: S1, constructing a BEV model; s2, feature grid mapping is carried out; s3, rendering a two-dimensional image; s4, performing self-supervised learning and optimization; s5, downstream task application; according to the method, BEV features are mapped into three-dimensional Gaussian parameters, end-to-end self-supervised learning is achieved through differential rendering, the method is applied to downstream tasks, and three-dimensional target detection, semantic segmentation or occupancy prediction are carried out; downstream tasks are trained and reasoned through self-supervision loss, no manual data annotation is needed, the data construction cost is remarkably reduced, and meanwhile the perception performance and generalization ability of the model in an automatic driving scene are improved.

The invention discloses an automatic detection and counting method for damaged and missing areas of a light-weight road waveform guardrail. The method comprises the following steps: S1, constructing a special data set; s2, carrying out image preprocessing and data annotation; s3, a lightweight GSYOLO model is constructed; s4, performing model training; s5, performing model evaluation; s6, guardrail damage and deficiency detection and counting analysis are carried out; and S7, outputting and visualizing a result. According to the method, automatic identification, accurate duplicate removal and quantifiable statistics of damaged sections and missing sections are realized, a structured detection result is output, and reliable and traceable data support is provided for road maintenance decision making.

The invention discloses an optical remote sensing ground feature relationship semantic understanding system and method in a localized environment, and the system constructs a remote sensing scene relationship analysis module, a multi-modal scene knowledge base module, a remote sensing scene graph construction module, a remote sensing scene representation module and a multi-modal sample pair construction module. Carrying out programmed modeling on the ground feature relationship by utilizing a code big language model adaptive to a domestic platform, and generating a scene graph triple; through a double-branch double-time phase comparison learning network, time sequence vision and structure knowledge are fused, and cross-modal joint representation learning is realized. According to the method, the problems of incomplete surface feature relationship expression, inaccurate modeling and lack of real-time semantics in the prior art are solved, deep optimization is carried out in the aspects of data annotation, heterogeneous computing power management and the like aiming at the localized environment, and the accuracy and integrity of semantic understanding of the surface feature relationship and the operation efficiency of the semantic understanding on a domestic software and hardware platform are remarkably improved.

The invention discloses an agent increment training system based on agricultural Internet of Things edge computing equipment operation feedback, and relates to the technical field of agricultural artificial intelligence decision optimization. Comprising an agent decision module, an execution module, a data acquisition module, an automatic labeling engine module and a model increment training module which are sequentially connected to form a closed loop. According to the method, the problems of automatic data annotation and confidence evaluation in agricultural decision model iteration are solved.

The invention relates to the field of electronic design automation, provides a method for semi-automatically processing an EDA (Electronic Design Automation) project file and generating a YOLO format annotation data set, and aims to solve the technical problems of low efficiency and poor precision of the traditional pure manual data annotation data set. According to the main scheme, the method comprises the steps of reading an engineering file of a jia-creative EDA project, analyzing and extracting element information, generating a standardized CSV data format, converting the standardized CSV data format into a label in a YOLO format, calculating normalized bounding box coordinates of elements, and recognizing special original elements and supplementing labeling information in combination with a template matching technology; drawing a bounding box on the schematic diagram image according to a YOLO label, highlighting different element types by using different colors, and generating a visual annotation image; revising and combining the newly generated YOLO label and a preliminary YOLO label generated by the AI model to obtain a label set; then importing into Labelimg for manual correction, and generating a final label set; and training a target detection model by using the final label set, and improving the pre-labeling precision through data set expansion iteration.

The invention relates to the technical field of medical image diagnosis, and discloses a radiation-pathological diagnosis evaluation method based on weak supervision cross-modal deep fusion. The method comprises the following steps: receiving case-level radiation image data and pathological section data, combining with a weak supervision consistency label, realizing cross-modal semantic alignment through a double-branch feature extraction network, and generating aligned radiation feature vectors and pathological feature vectors; based on the aligned feature vector, a cross-modal attention fusion mechanism is adopted to complete deep fusion, and a fusion feature vector is obtained; a consistency evaluation task is executed based on a multi-task learning framework, and a consistency classification result, an inconsistency attribution result and a risk area positioning result are output; and based on the evaluation result, generating a visual diagnosis report through an interpretability analysis model. According to the method, cross-modal data can be effectively fused under a weak supervision condition, the accuracy and interpretability of diagnosis consistency evaluation are improved, and clinical data annotation requirements are met.

The invention relates to the technical field of new energy project management, discloses a legal compliance file auditing method and system, and aims to convert a legal compliance file into a readable structured file through an optical character recognition method and solve the problem of high cost of traditional manual data annotation. According to the method, the file is preliminarily audited by using the preset multi-dimensional semantic recognition rule, so that the consistency of audit standards of different legal compliance files and different stages is ensured, and errors caused by artificial experience differences are avoided. Meanwhile, through multi-dimensional auditing, the accuracy of auditing is improved; and the preset intention classification model is used for carrying out deep semantic analysis on the file passing the preliminary auditing and carrying out secondary auditing, so that the auditing accuracy is improved. Therefore, by implementing the method and the device, a plurality of legal compliance files can be audited at one time, the manual audit workload of legal officers is reduced, the labor cost is reduced, the audit efficiency and accuracy are improved, and the compliance of legal compliance file management can be enhanced.

The invention discloses a crack detection model training method and device based on dynamic multi-strategy active learning, and the method comprises the steps: collecting and carrying out the preprocessing of a crack image, and constructing a data set containing a labeled subset and an unlabeled subset; performing preliminary training on the double-branch deep learning model by using the labeled subset; then, through an iterative active learning framework, comprehensively evaluating the value of an unlabeled sample from three dimensions of uncertainty, difficulty and representativeness, and dynamically adjusting the weight of each dimension according to a training stage to perform sample screening; and a mixed domain attention module fusing space and frequency domain information is introduced to realize more accurate difficulty perception, and the segmentation precision of the crack edge is improved by including boundary optimization loss. According to the method, the problems of high data labeling cost and low efficiency in deep learning crack detection can be solved, a detection model with better performance can be obtained with less labeling quantity, and the training efficiency and the detection accuracy are remarkably improved.

Techniques for masking data based on annotations are discussed herein. A system may receive a request to perform an action and leverage an LLM to assist in performing the requested action. When generating the input data to input to the LLM, the system can use a template to organize the input data. The template may include static data and / or slot(s) which can include a reference to data to input into such slots. The system may retrieve data to input to the slot based on the reference, retrieve annotations that define a classification of the data, and receive a policy that defines which types of data classifications are to be masked. Based on the data classification and the policy, the system can determine whether to mask the data. The system can generate the input data using the template, the data, and / or the mask(s) and input such data into the LLM.

One example method includes formulating a hypothesis for development of computing model, annotating the hypothesis with ethics metadata, storing the hypothesis and the ethics metadata, in association with each other, in a ledger, performing ‘n’ phases of a development lifecycle for the computing model, annotating each of the ‘n’ phases with ethics metadata specific to the phase, updating the ledger to include the ‘n’ phases and the ethics metadata respectively associated with each of the ‘n’ phases, and calculating an ethics confidence score for the computing model.

The invention relates to the technical field of industrial quality inspection, solves the technical problem that a traditional method cannot directly and effectively recognize a new scene due to insufficient generalization ability for the new scene which is not trained, and particularly relates to an industrial quality inspection method based on a multi-modal large model, which comprises the following steps: acquiring business terms contained in an industrial quality inspection scene to be applied; and a video sequence corresponding to the business terms; the video sequence is combined with cue words to be input into a multi-modal large model, and general description corresponding to the business terms is obtained; obtaining N frames of to-be-identified video sequences; and inputting the video sequence, the business terms, the general description and the cue words into a multi-modal large model which is subjected to fine tuning training in advance to obtain an industrial quality inspection result. According to the method, the multi-modal large model can further complete business term alignment on the basis of general description of knowledge of the multi-modal large model, any industrial quality inspection scene task can be supported without a large amount of data annotation, and the generalization ability of a new scene is greatly improved.

The invention discloses a multi-modal heterogeneous knowledge base automatic construction and evolution updating method and system, and belongs to the technical field of artificial intelligence and knowledge base data processing, and the method comprises the following steps: S1, multi-modal data collection and cross-modal semantic alignment; s2, human-social multi-source data labeling and information quality screening are carried out; s3, constructing and dynamically updating a multi-modal heterogeneous knowledge base; and S4, system integration and packaging service: packaging the above processes into an integrated multi-mode knowledge base construction and updating system. Unified representation, high-quality knowledge precipitation and dynamic evolution of multi-source heterogeneous data in the human society field can be realized, and a high-quality, reasonable and high-timeliness knowledge infrastructure is provided for human society intelligent service.

The application discloses a kind of based on data center application inference big model discovers network topology fault and false alarm self-healing visualization method and system thereof, including using grid division algorithm carries out grid division and forms DeepSeek inference network topology and chip resource mark;Visual image processing is carried out in DeepSeek inference network topology, complete network topology resource re-planning and visual data annotation, output DeepSeek inference big model inference conclusion;From network topology gateway configuration file, the transmission protocol of application is used as input parameter and carried out Deepseek inference big model inference analysis, carries out power quantization;DeepSeek inference network topology node is associated, and alarm is carried out in transmission protocol through the fluctuation of power level.The application uses microfluidic chip to manage the chip that inference big model carries out operation, chip resource mark is input into inference big model, and then uses image definition position and inference big model embedded AI algorithm matching discovers network topology, improves big model inference conclusion and visual accuracy.

This application discloses a data annotation method, electronic device, and storage medium. The method includes: annotating raw sensor data collected by multiple sensor devices to obtain several initial bounding box information corresponding to the same target; fusing the several initial bounding box information in a bird's-eye view coordinate system to obtain fused bounding box information of the target; and updating the initial bounding box information of the target in each raw sensor data based on adjustment information of the fused bounding box information. This approach can improve the efficiency and accuracy of image data annotation.

The invention discloses a data portraying method based on high-quality data elements. The method comprises the following steps: acquiring source information of data from an acquisition system through a data source recording module; acquiring a data processing historical record from the data medium table through a processing process recording module; acquiring data annotation related information from an annotation system through an annotation method recording module; performing quality verification on the data elements through a verification tool module; and recording and storing all verification results through a verification result recording module. According to the invention, through cooperative work of the five modules, a complete portrait of the whole life cycle of the data elements is realized, the normalization, transparency and credibility of data management are effectively improved, and a reliable data basis is provided for data-driven business decision.

The invention relates to a method for data annotation (10) of sensor measurement data (M) from environmental sensors (S) of a vehicle (12), comprising the steps of providing at least first sensor measurement data (M1) of a first environmental sensor (S1) and second sensor measurement data (M2) of a second environmental sensor (S2) to a common environmental scene (14) of a vehicle environment (20) of the vehicle (12) comprising at least one environmental object (16), creating (26) a first annotation basis (A1) comprising at least the first and second sensor measurement data (M1, M2) combined in a common first time period (28), annotating (42) the first sensor measurement data (M1) of the first time period (28) to identify environmental objects (16), and projecting (46) the annotation (42) of the first sensor measurement data (M1) onto the second sensor measurement data (M2) of the first time period (28).Verification (62) of the annotation (42) of the first sensor measurement data (M1), at least taking into account the projection (46) into the second sensor measurement data (M2), and output (60) of the verified annotated first sensor measurement data (M1). Furthermore, the invention relates to a method for generating model data.

This invention discloses a system and method for labeling automotive abnormal noise data, relating to the field of automotive NVH detection technology. The method includes: receiving raw signal data of automotive abnormal noise; performing time-frequency transformation on the raw signal to generate a time-frequency diagram; identifying key frequency ranges characterizing the abnormal noise features; then performing directional filtering to enhance the abnormal noise signal; identifying candidate time ranges for abnormal noise through an automatic detection algorithm; and outputting the final abnormal noise time range labels and key frequency ranges in a structured format after verification and correction. The system includes a data receiving module, a signal processing module, a display module, an intelligent labeling module, an audio playback module, a labeling output module, and an interaction module, achieving triple-assisted labeling through auditory, visual, and intelligent pre-selection methods. This solves the problems of low efficiency, low accuracy, and poor consistency in traditional manual listening methods, providing high-quality labeled data for deep learning classification of automotive abnormal noises.

This invention relates to a method for Gaussian point cloud scene model extraction and 3D semantic segmentation. It solves the problems of low quality, poor accuracy, and lack of detail in existing image 3D processing techniques. The method includes: S1, reading the image and obtaining initial Gaussian point cloud data; S2, inputting an RGB image and outputting it as a data annotation module; S3, calculating the distance from each pixel to the nearest background point; S4, randomly selecting a viewpoint in the differentiable rendering module, projecting the Gaussian point cloud from the world coordinate system to the image coordinate system using a 2DGS projector, and rendering the image from that viewpoint using the 2DGS renderer; S5, training the target using boundary loss; and S6, testing on a public dataset and a self-made dataset, presenting and explaining the results in two parts. The advantages of this invention are: fast training speed, high segmentation accuracy, convenient operation, and effective improvement of image rendering quality.

The invention relates to a fault report analysis system and method based on image recognition, and the method comprises an image collection module which calls a camera API to obtain a fault report image; the image recognition and data extraction module is used for recognizing character information in the fault report image based on a deep learning OCR technology so as to extract a fault current value and distance measurement data; identifying and positioning a graph area based on a computer vision algorithm, and identifying data annotations in the graph in combination with template matching or a deep learning target detection model; the data processing and calculating module is used for arranging the fault current values, the distance measurement data and the data marks in the graph in an ascending order or a descending order according to timestamps; zero-sequence current is determined according to the fault current value, the ranging data and the data mark in the graph; and the report generation module is used for designing a fixed template and filling corresponding data in corresponding positions of the template to generate a final fault report. The method has the beneficial effects that the acquisition mode of the zero-sequence current data is optimized, and the intelligent level of fault report analysis and report is improved.

The invention belongs to the technical field of data processing, and relates to a data annotation processing method and device, computer equipment and a storage medium, and the method comprises the steps: screening to-be-annotated specified data from a data set; performing content configuration on the labeling template based on the specified data to obtain a target labeling template; performing data rendering on the target annotation template based on the specified data to generate a target annotation page; creating an annotation task corresponding to the target annotation page, and distributing the annotation task to annotation personnel based on the task distribution mode; receiving a generated annotation result corresponding to the target annotation page; carrying out quality inspection processing on the labeling result based on a quality inspection strategy; if the labeling result passes the quality inspection, carrying out acceptance check processing on the labeling result based on an acceptance check rule; and if the labeling result passes the acceptance check, performing output processing on the labeling result. The method can be applied to data annotation scenes in the fields of finance, science and technology and medical health, and the processing efficiency, accuracy and annotation quality of data annotation processing can be effectively improved.

An adaptable deep learning method is provided that delivers sound hepatic lesion identification in NETs, while significantly reducing human effort for data annotation and improving model generalizability for PET image quantification. A region-guided GAN (RGGAN) model conducts image-to-image translation between list-mode simulated PET images and real-world clinical data, while preserving semantic content of interest, e.g., lesions. The RG-GAN model is integrated with a lesion detection model into an end-to-end, unified framework for joint-task learning, such that the two models can benefit from each other. The RG-GAN translates the list-mode simulated data into real world-style images, which appear to be drawn from the real clinical PET image dataset, and feeds the translated images into the lesion detection model for training. In order to deal with the limited diversity of list mode-simulated PET image data, a specific data augmentation module is incorporated into the unified framework to improve model training.

The invention relates to the technical field of data annotation, and discloses an annotation result processing method and device and electronic equipment. According to the method and the device, invisible annotation objects in the annotation result are removed, the problem of single-frame annotation noise can be avoided, the quality of a network training result can be improved, and illusion is avoided; a plurality of observation visual angles can be integrated based on ray casting, and more comprehensive target information is obtained through ray tracing, so that the sensing ability of a small target is enhanced, and the detection recall rate is improved; the method can also be applied to pure visual scenes. In addition, calculation can be directly carried out at the user side by using a rendering engine without depending on the communication and computing power support of the cloud side, so that the calculation can be carried out in real time in the labeling and quality inspection stages. The scheme of the invention has great advantages in the aspects of applicability, accuracy, real-time performance and the like.

Embodiments of the present disclosure disclose a graph construction method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises: in response to receiving a user's question, obtaining a graph construction request text according to a keyword extracted from the user's question; performing format processing on the target business graph construction request text to obtain a processed graph construction request; generating a basic target business graph based on the processed graph construction request and a preset large language model; generating a basic target business graph supplement text set based on the target business related content of the text vector library and the basic target business graph; extracting the basic target business graph supplement text set to obtain a supplement triple; and generating a complete target business graph by combining the supplement triple and the basic target business graph. This embodiment can generate a graph with a wider coverage, eliminates the data annotation process, and can connect corresponding devices to operate the graph, thereby improving user experience.

The invention provides a defect image clustering method, a defect detection system and a medium, and the method comprises the steps: obtaining a defect image clustering model through two-stage training, and based on the clustering model, according to one or more to-be-detected defect region images of a to-be-detected sample and the surface feature data of each to-be-detected defect region image, obtaining a defect image clustering model; performing clustering analysis operation on each to-be-detected defect area image, and predicting a clustering result of each to-be-detected defect area image; therefore, the defect type of each to-be-detected defect area image can be effectively recognized, the method has a remarkable effect on recognition of complex defects, the clustering precision is high, defect problem qualification and process abnormality root positioning are facilitated, and continuous improvement of a semiconductor or generic semiconductor manufacturing process is guided; and each clustering result can be used for training a subsequent defect image classification model, so that the time consumption of data annotation can be effectively reduced, and the labor cost is saved.

The invention discloses a trusted space-based data annotation management and control method, system and device, and a medium. The method comprises the following steps: responding to a data labeling task request, determining a task type and labeling a task life cycle, dividing the task into a plurality of finite states, and binding a dynamic strategy set for each state; determining the computing power demand of each state according to the task type, and obtaining and configuring a corresponding resource template; when a user initiates a trusted spatial data loading request, obtaining user information and a loading demand, activating a resource template in an authentication state, and performing authentication according to a dynamic strategy set; after the authentication is passed, triggering a state transition event to enter a data loading state, and loading target data; after labeling is completed, a labeling result is obtained and subjected to quality inspection auditing; and if the audit is passed, outputting a result, and uploading the migration event and the annotation result for evidence storage. According to the method, the labeling work can be efficiently and safely completed according to the task, meanwhile, the integrity, traceability and safety of the data are effectively guaranteed, and the quality and efficiency of the labeling work are improved.

The invention relates to the technical field of image processing, and discloses a data annotation method and device, a storage medium and data annotation device.The data annotation method comprises the steps that to-be-annotated data and a trained target model are obtained; preprocessing the to-be-labeled data through a 3D point cloud cross-mode and a preset model to obtain structured data of the to-be-labeled data; inputting the structured data into the target model to determine a semantic mask of the to-be-labeled data; obtaining an uncertainty score of the semantic mask; and when the uncertainty score is greater than a preset score, updating the target model, and performing data annotation by using the updated target model. Therefore, the precision and integrity of semantic annotation in a complex scene are remarkably improved, the model can be dynamically optimized, and manual intervention can be reduced.

The application discloses a kind of method and system for big data acquisition and annotation, it is related to image annotation technical field.The method comprises: according to the Q edge monitoring nodes of monitoring system, obtains Q real-time monitoring set;According to image quality analysis factor, image quality annotation learning is carried out to Q edge monitoring nodes, and acquisition quality annotation channel is constructed;Q real-time monitoring set is adaptively reinforced, and Q reinforcement monitoring set is obtained;Introduce differentiating attention learning mechanism, and carry out annotation feature federation learning to Q edge monitoring nodes, and obtain edge annotation multi-channel;Introduce annotation loss iterative optimization mechanism to carry out cloud distillation to edge annotation multi-channel, and obtain cloud annotation multi-channel;According to cloud annotation multi-channel, Q reinforcement monitoring set is automatically annotated.Solve the technical problems of low efficiency and insufficient accuracy of monitoring image data annotation in the prior art, and achieve the technical effect of improving image annotation efficiency and accuracy.

This application provides a method, system, medium, and terminal for constructing a knowledge graph of faults in key ship equipment. The technical solution of this application is aimed at building a professional ontology model for ship equipment fault diagnosis, defining seven types of entities and five types of relationships, and is compatible with component nesting and bilingual expression. It can completely depict the equipment architecture, fault propagation, and maintenance knowledge. Relying on the large model prompt word engineering, it achieves efficient knowledge extraction with few samples, saving large-scale data annotation and reducing the cost of database construction. It integrates a hybrid alignment algorithm of semantic vectors and edit distance, coupled with a three-level decision mechanism, to efficiently solve the problem of differences in domain entity representation and complete the fusion of multi-source knowledge. The overall solution has outstanding versatility and scalability, and can be quickly adapted to the construction of fault knowledge graphs for various industrial equipment by fine-tuning parameters.

This invention relates to the field of assisted driving technology, specifically to a method for detecting pole-shaped obstacles in assisted driving, comprising the following steps: creating a self-made pole-shaped obstacle dataset; labeling the acquired images using data annotation software, and processing the labeled files to divide them into training, validation, and test sets according to proportions; using an improved loss function based on the YOLOv5s network to backfeed network parameters; introducing the MK function into the original loss function; adding a dual-channel hybrid attention mechanism (CBAM) after the residual module; introducing an improved multi-scale feature fusion network structure into the neck network; training the model; and applying the trained model to pole-shaped obstacle detection. This method not only improves the recall and precision for detecting obstacles in the complex and ever-changing traffic environment of urban roads, but also solves the problem of variations in the size and shape of pole-shaped obstacles at different scales.