893 results about "Data labeling" 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 discloses a multi-modal data labeling method and system based on large model pre-labeling, and the method comprises the steps: S1, receiving to-be-labeled multi-modal original data and labeling task definition, and generating a structured task instruction signal; s2, inputting the structured task instruction signal into a multi-modal large model, and generating a pre-labeling result signal containing a preliminary label and a corresponding confidence coefficient thereof; s3, scheduling a manual verification task based on the confidence coefficient in the pre-labeled result signal; s4, according to the manual verification signal, performing parameter fine tuning or prompt optimization on the multi-modal large model, and generating a model optimization signal; and S5, pre-labeling the new multi-modal original data by using the multi-modal large model updated by the model optimization signal, and fusing an artificial verification signal. According to the multi-modal data annotation method and system based on large model pre-annotation, the problems that traditional multi-modal data annotation is low in efficiency, high in cost and difficult to unify in quality can be solved.

The invention discloses an information extraction data annotation and training system based on a large language model, and relates to the technical field of information extraction, and the system comprises a data set construction module which is used for constructing a pre-training data set and a fine tuning data set; the model continuous pre-training module is used for carrying out continuous pre-training on a preset general large language model based on the pre-training data set to generate a field adaptive pre-training model; the model fine tuning module is used for performing supervised fine tuning training on the domain adaptive pre-training model through a two-stage course learning strategy based on the fine tuning data set, and generating an information extraction model; the retrieval enhancement generation module is used for performing entity-semantic retrieval on an input text based on a preset knowledge base, outputting context information related to the input text, and outputting structured information of the input text based on the context information and an information extraction model, the problems of insufficient generalization ability, poor field adaptability and disastrous forgetting of a general large language model are solved, and the accuracy and robustness of information extraction are improved.

The embodiment of the invention provides an annotation task allocation method and device, equipment, a storage medium and a program product. The method comprises the following steps: acquiring a plurality of annotation tasks to be distributed; determining task portraits of the plurality of to-be-distributed annotation tasks according to the plurality of to-be-distributed annotation tasks; obtaining capability portraits of a plurality of annotation personnel and capability portraits of a plurality of data annotation models, wherein the data annotation models are used for data annotation; according to the task portrait, the ability portraits of the plurality of annotation personnel and the ability portraits of the plurality of data annotation models, constructing an optimization model used for determining an allocation relationship among the to-be-allocated annotation task, the annotation personnel and the data annotation models; and solving the optimization model by using a solving algorithm to generate a target task allocation scheme.

The invention provides a data labeling method and system, and the method comprises the steps: carrying out the voxelization processing of a laser radar point cloud and a three-dimensional image of a target scene, and obtaining multi-scale voxel features; performing sparse interaction on the point cloud voxel features and the image voxel features of the same scale to generate multi-modal sparse interaction voxel features; determining a three-dimensional bounding box of at least one target object in the target scene and a corresponding confidence coefficient; according to a natural language instruction provided by a user and used for describing a target object, generating a description text corresponding to the natural language instruction through the large language model; driving a preset prompt type visual model based on the description text, and generating a two-dimensional segmentation mask corresponding to the target object through an iterative interaction mode; according to the method, the three-dimensional bounding box is matched with the two-dimensional segmentation mask, and the semantic category represented by the two-dimensional segmentation mask is labeled to the matched three-dimensional bounding box, so that the manual intervention frequency is reduced, and the labeling efficiency, flexibility and consistency are improved.

The invention belongs to the technical field of data processing, and provides a data labeling method and system based on cue word driving, a joint guide vector is generated by fusing field features of a labeling task and operation behavior vectors of labeling personnel, labeling errors caused by guide deviation are greatly reduced, and labeling efficiency and preliminary labeling quality are improved; the initial labeling result is analyzed, the domain distribution difference between the labeling defect type and the associated cue word is recognized, and clear targeting is provided for follow-up knowledge graph parameter optimization; a reward function is constructed by taking a labeling defect type and cue word field distribution difference as a state space and combining labeling accuracy and field adaptability, parameters of a cue word knowledge graph are iteratively corrected through reinforcement learning, and high-quality cue words can be continuously output; after the cue word knowledge graph iteration is stable, the fusion coefficient of the joint guide vector is updated based on quality evaluation data feedback, and it is ensured that the joint guide vector and the optimized knowledge graph are cooperatively matched.

The invention relates to the technical field of training and optimization of a large language model, discloses a direct preference optimization method and device based on importance sampling, and effectively solves the problem of low optimization efficiency caused by overall processing of a response sequence in an existing direct preference optimization method by introducing a character marking level importance sampling mechanism. According to the method, the importance weight of each character mark is estimated by utilizing the positive preference model and the negative preference model, and the optimization target is weighted, so that the training resources are concentrated on the key character marks, and therefore, on the premise that the data marking cost is not increased and the original concise process is not changed, the data marking efficiency is improved. More efficient and more stable model preference alignment with good interpretability is realized, and the optimization efficiency and the output quality of the large language model are remarkably improved.

The invention discloses a mass data automatic label generation method based on a large model and a rule engine, which comprises the following steps: S1, multi-source data acquisition and preprocessing: acquiring data through a unified data interface, and then performing data cleaning, data standardization processing and data format conversion to obtain standard input data; s2, constructing a label system framework; s3, generating a large model label; s4, performing intelligent clustering and label refining; s5, rule engine constraint and optimization; s6, evaluating and optimizing label quality; and S7, automatically expanding and updating the tag system. According to the method, automatic label generation of mass multi-source heterogeneous data is realized through a large model and rule engine technology under the condition of no personnel intervention, a large-scale label system containing thousands of labels is constructed, meanwhile, the label system can be rapidly updated along with product iteration and keep timeliness, and the generated labels are accurate and extensible.

The invention relates to the technical field of data annotation, and discloses a data annotation system and method for an automatic driving scene, and the method comprises the steps: constructing a scene database and a data set for managing to-be-annotated data, working condition data and accepted data; creating an annotation item and obtaining to-be-annotated data; the to-be-labeled data and the corresponding working condition data are associated, scene criticality evaluation is carried out, a scene criticality evaluation driving mechanism is introduced to carry out dynamic strategy generation, and a task strategy and a quality inspection strategy are included; splitting the to-be-labeled data according to a task strategy and performing task allocation; performing intelligent labeling model training based on the data set corresponding to the to-be-labeled data category; pre-labeling by utilizing the trained intelligent labeling model, carrying out manual labeling by utilizing a plurality of built-in labeling tools, and outputting labeled data; performing quality inspection on the labeled data according to a quality inspection strategy so as to output checked and accepted data; and performing intelligent labeling model training and updating based on the checked and accepted data. According to the invention, a high-expansibility system is constructed, and man-machine cooperation high-quality labeling is realized.

The invention discloses a federal category forgetting method based on label reconstruction and collaborative compression. The federal category forgetting method comprises the implementation steps that a client performs gradient matching optimization on samples randomly sampled from original data to obtain compressed data; after a forgetting request is received, reconstructing a local data label needing to be forgotten, and combining the data label with non-forgetting compressed data to form a forgetting training set; freezing the original model, finely adjusting an additional low-rank matrix by using a forgotten data set, regularizing by applying an F2 norm, and uploading the trained low-rank matrix to a server; and the server aggregates the low-rank matrix uploaded by the client and performs sparse processing, and then sends the low-rank matrix back to the client for training. According to the method, forgetting of forgotten category data is realized while remaining category performance of the model is kept, and calculation and communication overhead in a federal category forgetting process is greatly reduced.

The invention relates to the technical field of data processing, in particular to an intelligent matching method, device, equipment and medium for achieving hospital value domain standard mapping, and the method comprises the steps: importing hospital standard value domain data into a standard library, and importing to-be-matched hospital original value domain data into a matching database, performing standardized processing on the data, and performing classified storage according to data categories; performing mapping matching on the hospital original value domain data to be matched by adopting a multi-level matching strategy, including precise matching, fuzzy matching, empirical value matching and AI matching which are executed in sequence, and calculating similarity scores of matching levels respectively; checking a matching result, confirming a correct matching result and storing the correct matching result as empirical value data; and marking the data which are wrongly matched in the auditing as negative samples, marking the corrected correct data as positive samples, forming a training set after data marking, and carrying out iterative training optimization on the Rianker algorithm model. And the accuracy of matching results is improved.

The invention relates to the technical field of computer data processing, and discloses a multi-modal data labeling method based on a large model technology, and the method comprises the steps: 1, carrying out the data collection of multi-modal data, and carrying out the data preprocessing of the collected multi-modal data; 2, single-mode feature extraction is carried out on each kind of single-mode data in the multi-mode data, cross-mode feature fusion is carried out, and a fusion feature vector capable of being co-processed is generated; 3, performing model training according to the fusion feature vector to obtain a joint training model, and labeling the multi-modal data based on the joint training model to obtain a preliminary labeling result; and 4, carrying out manual intervention and data storage on the preliminary labeling result, and outputting a labeling result. By means of the scheme, accurate and efficient annotation of various types of data such as texts, images, voices and videos can be achieved, and deep application and development of multi-modal data in the field of artificial intelligence are promoted.

The invention discloses a data annotation agent network attack identification method based on artificial intelligence. The method comprises the following steps: collecting and preprocessing original network monitoring data; constructing a labeling task set; performing feature analysis, generating an initial category label and a corresponding confidence value, and outputting an agent labeling result; according to the confidence value of each agent, a preset weight parameter and an annotation consistency index, generating an annotation sample set; dividing the labeled sample set into a network attack identification training sample set and a network attack identification verification sample set; the improved HTAN model is trained, and a trained improved HTAN model is obtained; outputting an attack probability value and an attack category judgment result of the online detection sample; according to the method, efficient labeling, high-precision learning and real-time detection of network attack identification are realized, and the automation degree and attack identification capability of a network security protection system are improved.

The invention relates to the technical field of speech synthesis, in particular to a speech synthesis-based corpus expansion method, system and device and a medium, and the method comprises the steps: carrying out the preprocessing including data annotation based on the collected audio and corresponding text of a target speaker; extracting acoustic features from the preprocessed audio; on the basis of a pre-trained acoustic model, performing personalized fine tuning by using the annotation data and the acoustic features, and training personalized acoustic models of a plurality of speakers at the same time through multi-thread parallel computing; calling the trained personalized acoustic model, and synthesizing a voice corpus of the target text in combination with a vocoder; and based on the trained personalized acoustic model, continuously expanding the corpus by changing the text. The personalized voice corpus is quickly generated through a small number of voice samples, the data acquisition cost is remarkably reduced, and the corpus construction efficiency is improved.

The invention discloses a vehicle fine granularity detection method based on a three-dimensional grid and YOLOv11 transfer learning. The method mainly comprises three components: a source domain prediction network structure, a target domain prediction network structure and a transfer learning module. Wherein the source domain prediction network and the target domain prediction network are dual-channel deep networks fusing two-dimensional images and three-dimensional grids, and efficient migration of source domain knowledge in a target domain is realized by aligning feature distribution of the source domain and the target domain. Through deep fusion of three-dimensional grid information and two-dimensional image features, the method can maintain robust detection performance under adverse conditions of vehicle attitude change, illumination interference, shielding and the like, can effectively reduce large-scale data annotation and training cost, improves the rapid adaptation capability of the model in a new scene or a new vehicle type, and improves the robustness of the model. And a high-precision, extensible and rapid-iteration fine-grained identification and detection solution is provided for intelligent traffic monitoring, unmanned driving perception, military equipment identification and digital twin systems.

The invention discloses a zero-sample multi-modal relation extraction method based on a multi-modal large model, which comprises the following steps of: constructing prototype information containing tag names, descriptions and aliases for seen and unseen relation categories, and encoding and aggregating the prototype information into prototype vectors; extracting feature representation of a training sample through a multi-modal large model, and performing fine adjustment on the model by updating low-rank adapter parameters based on the feature representation and a known category prototype vector; and for the input containing the unseen category, extracting the features of the input by using the fine-tuned model, and completing relation identification in combination with the prototype vector of the unseen category. According to the method, the structured prototype knowledge is injected into the low-rank fine tuning process, so that the model keeps semantic perception of the unseen relationship while absorbing the domain knowledge, the relationship extraction accuracy and generalization ability in a zero sample scene are remarkably improved, and the data annotation cost is effectively reduced.

The invention discloses an oxidation ditch sludge state monitoring method and system, and relates to the field of computers, and the method comprises the steps: obtaining image data; generating unique metadata for each piece of image data; performing data annotation on the image data; recognizing the image data according to the image recognition model to obtain a recognition result; the metadata and the recognition result are fused, and cue words are constructed; inputting the cue word into the large model to obtain an analysis result; and generating a diagnosis report according to the analysis result. Through high-precision image monitoring and the deep reasoning analysis capability of a large model, the sludge state change is mastered in real time, so that the real-time, accurate and intelligent monitoring and diagnosis of key indexes such as sludge structure stability and filamentous bacteria abundance are realized.

The invention discloses a young poultry counting method based on deep learning, and relates to the field of image processing, and the method comprises the steps: information collection and feature extraction: collecting a young poultry image through a hardware platform, and carrying out the preprocessing of the image, and obtaining a continuous color image of a target image; data annotation: carrying out label annotation on the preprocessed image, carrying out accurate target detection annotation on stacked young poultry in the image to obtain an annotated data set, and providing a high-quality data set for model training; inputting the collected real-time image into a young poultry detection system constructed based on a YOLO architecture, wherein the system is used for performing target detection, target tracking and target counting on the input image; and multi-target tracking: continuously positioning spatial positions of a plurality of targets through frame-by-frame analysis of the video sequence, and maintaining a unique identity (ID) of each target. Precise counting and high-speed real-time processing of the young poultry in a high-density scene are realized.

The invention discloses a load identification method and system based on multi-view learning and a deep wide residual network. The method comprises the following steps: acquiring voltage and current data at an outlet of a current collector; extracting two types of load feature views from the collected data, and adding data labels to form a labeled data set; generating two training subsets from the labeled data set, wherein the initial training subset comprises a part of labeled samples and unlabeled samples; performing cross training on the two depth wide residual classification networks by using the training subset, and training to obtain a depth wide residual classification network model based on multi-view learning; and extracting two feature views from unknown electric appliance data collected in real time, inputting the two feature views as label-free data into the depth wide residual classification network model based on multi-view learning, updating the model, and obtaining a load identification result. The method has the advantages of improving the model recognition accuracy and generalization performance, enhancing the robustness of the model, improving the model recognition accuracy and saving the model training time.

Embodiments of the present disclosure generally relate to methods for autonomous orchestration of a data labeling workflow. Embodiments include generating, using a generative machine learning model, a natural language description of input data. Embodiments include identifying candidate labels that are semantically similar to the natural language description of the input data. Embodiments include providing natural language descriptions of the candidate labels and the natural language description of the input data to a language processing machine learning model. Embodiments include receiving an output from the language processing machine learning model in response to the natural language descriptions of the candidate labels and the natural language description of the input data, wherein the output indicates a selected label from the candidate labels. Embodiments include validating the output based on an alternative label determination technique. Embodiments include associating the selected label with the input data based on the validating.

The invention relates to the technical field of data processing, and discloses an automobile fault code labeling method and device, computer equipment and a storage medium, and the automobile fault code labeling method comprises the steps: carrying out the recognition processing of fault maintenance data, generating structured data, carrying out the standardization processing of the structured data, and generating text verification data; fault code fields and state information are extracted based on the text verification data, preliminary annotation data are constructed, recognition enhancement processing is carried out according to the preliminary annotation data, and fusion annotation data are generated; carrying out maintenance part matching processing on the fusion annotation data to generate image-text annotation structure data; and performing format processing on the image-text annotation structure data to generate target automobile fault code annotation data. According to the method, the labeling capability of the complex maintenance data can be effectively improved, the manual operation burden is reduced, and the efficiency, accuracy and consistency of data labeling are greatly improved.

The invention relates to a sleep monitoring and early warning method and system based on respiration data analysis and a medium, and belongs to the technical field of .The respiration data of a user is predicted through a sleep state and sleep apnea hypopnea index prediction model, the sleep state of the user and the sleep apnea hypopnea index are estimated, and the sleep apnea hypopnea index is obtained. And finally, early warning is performed according to the sleep state of the user and the sleep apnea hypopnea index, and meanwhile, a related treatment scheme is generated according to early warning information. According to the method, a deep learning model is pre-trained, sleep staging and sleep apnea hypopnea index estimation are carried out on the model in a unified framework at the same time, and internal correlation between a sleep macrostructure and a respiratory event is effectively decoupled. Then, through a domain adversarial training mechanism, sleep respiration characteristic knowledge learned in the contact type respiration signals is migrated to millimeter wave radar signals, and the problem that the generalization ability of a model is insufficient due to scarcity of radar data labels is solved;

To provide a technique capable of more accurately associating an article with an identification label attached to the article and improving the tracking accuracy of the article.SOLUTION: An imaging unit 4 that acquires image data obtained by imaging the article 2 being conveyed; a label recognition unit 11 ba that detects the identification label 3 in the image data and acquires position information of the detected identification label 3 in the image data as label position information; The article management device includes an article recognition part 11Bb for acquiring position information in the image of the detected article 2 as article 2 position information, and an association part 11C for associating the article 2 with the identification label 3 from the label position information and the article 2 position information acquired from the same image.SELECTED DRAWING: Figure 4

The invention discloses an intelligent sand table identification system based on an artificial intelligence multi-modal technology, which belongs to the field of sand table identification and comprises a multi-modal data acquisition module, a multi-modal feature fusion module, an intelligent analysis and interaction module and a system integration optimization module. According to the method, the problems of insufficient recognition robustness, poor dynamic analysis real-time performance, weak cross-scene generalization ability, high data labeling cost and single interaction mode caused by dependence on a single sensor and lack of multi-dimensional perception ability in the prior art are solved, the limitation of traditional single-mode recognition is broken through by collecting vision, force sense, voice and environment data, and the recognition efficiency is improved. The real-time intelligent analysis and dynamic interaction technology is adopted, real-time intelligent analysis and dynamic interaction of sand table operation are achieved, the system further has strong generalization ability, a meta-learning framework and an active learning strategy are adopted, dependence on annotation data is reduced, the system is reasonable in architecture, distributed collection, heterogeneous calculation and micro-service deployment are adopted, and the system has good application prospects. And high-efficiency operation and expandability are ensured.

The invention provides a threat detection method and device, and the method comprises the steps: constructing a time sequence diagram sequence corresponding to each time window through the preprocessed multi-source log data of a target network; embedding nodes in the time sequence diagram sequence corresponding to each time window by using a relational graph neural network to obtain an embedded sequence matrix; inputting the embedded sequence matrix into a Transform model, and performing autoregressive prediction to obtain a node prediction matrix of a next sub-graph node; constructing an attention score matrix through the predicted node prediction matrix of the next sub-graph node, so as to predict a connection matrix at the next moment; and calculating an error between the connection matrix at the next moment and the real adjacent matrix at the next moment to determine whether a security threat exists at the next moment. According to the scheme, the technical problems of data imbalance and data annotation in the prior art are solved, and efficient identification of low-frequency and hidden attack behaviors is achieved.

The invention discloses a time sequence physiological signal data generation method and system, a terminal and a storage medium, and relates to the technical field of data processing.The method comprises the steps that a reference time sequence physiological signal, a reference data label and a control data label corresponding to a target object are obtained, and the reference data label is matched with the reference time sequence physiological signal; performing individual feature extraction according to the reference time sequence physiological signal and the reference data label to obtain an individual feature representation vector corresponding to the target object; according to the individual feature representation vector and the control data label, a target time sequence physiological signal corresponding to the target object is generated through a trained conditional diffusion generation model, and the target time sequence physiological signal is matched with the control data label. In this way, automatic generation of the time sequence physiological signal data can be achieved, and therefore sufficient time sequence physiological signal data can be efficiently provided.

The invention relates to the technical field of data annotation quality evaluation, in particular to a dynamic quality evaluation method and system based on multi-modal enhancement and active learning. The method comprises the steps of calculating a distribution difference between annotation data and historical annotation data, generating a data drift judgment result and outputting a data drift perception quality signal; calculating a consistency score based on a modal confidence consistency function, forming a cross-modal consistency evaluation quality signal, and combining the cross-modal consistency evaluation quality signal with the data drift perception quality signal to form a multi-dimensional quality signal vector; analyzing the risk level of the annotated data by adopting a dynamic risk grading active learning method, and adjusting a sampling inspection proportion and a release threshold; and constructing a label ambiguity scoring function in combination with the cross-modal consistency evaluation quality signal, calculating a label ambiguity score value of the labeled data, and judging whether the labeled data is in a reasonable label divergence range. According to the method, the dynamic evaluation and optimization of the annotation data quality are realized, and the quality control of the data annotation process is improved.

The present disclosure relates to the technical field of data processing, and disclosed are a multi-modal generic perception model training method and apparatus, a labeling method and apparatus, and an electronic device. The method comprises: determining a plurality of sample images, labeling each sample image, and determining a plurality of initial region-level image-text data labeling results corresponding to each sample image; correcting the plurality of initial region-level image-text data labeling results corresponding to each sample image, and determining a plurality of corrected labeling results corresponding to each sample image; and on the basis of the plurality of corrected labeling results corresponding to each sample image, training a multi-modal generic perception model to obtain a trained multi-modal generic perception model, wherein the trained multi-modal generic perception model is used for labeling a target image and determining a plurality of region-level image-text data labeling results corresponding to the target image.

The embodiment of the invention provides a data labeling method and device, a vehicle, a storage medium and a program product, and the method comprises the steps: obtaining modal data in at least one modal collected by a plurality of sensors disposed in the vehicle, and obtaining multi-modal data; the multi-modal data is input into an annotation model for annotation, at least one initial annotation result is obtained, the annotation model is obtained through training based on a multi-modal data sample collected by a sensor and an annotation result sample corresponding to the multi-modal data sample, and at least one annotation result is obtained; the initial labeling result is at least used for representing the position of at least one obstacle in the traffic area; in response to the fact that the confidence coefficient of the initial labeling result is lower than a confidence coefficient threshold value, obtaining an adjustment instruction; and in response to the adjustment instruction, adjusting the three-dimensional space label of the obstacle in the initial labeling result to obtain a target labeling result. The technical problem of low accuracy of data processing is solved.

The invention discloses a multi-modal sensing data automatic labeling method and system for gait assessment, and the system comprises a sensing collection module which comprises two paths of pressure sensors and a six-axis IMU sensor; the two paths of pressure sensors are used for acquiring plantar pressure data; the six-axis IMU sensor is used for acquiring lower limb movement data; the wireless transmission module is used for uploading the collected plantar pressure data and lower limb movement data to a cloud Internet of Things platform; the cloud data processing module is used for receiving and storing the plantar pressure data and the lower limb movement data; the data labeling module is used for automatically recognizing a gait cycle and carrying out multi-dimensional score labeling; and the user interaction module is used for interacting with the data annotation module and the cloud data processing module and displaying gait evaluation related data and results. According to the invention, a multi-mode sensing PCB is self-developed, a closed-loop system of collection, cloud loading, labeling and display is constructed, and accurate, continuous and remote evaluation of gaits of children with cerebral palsy is realized.