113 results about "Label quality" patented technology

The invention discloses a long document knowledge base construction method based on directory perception and a program product, and belongs to the field of artificial intelligence and natural language processing. According to the scheme, an original long document is sequentially subjected to preprocessing, directory structure analysis, mixed blocking, double-tag generation, tag intelligent optimization, vectorization and meta-information mounting, and finally automatic construction and high-quality retrieval enhancement generation of a knowledge base are achieved. According to the method, the semantic integrity is guaranteed by fully utilizing the perceptual ability of the directory structure, the label quality and the retrieval efficiency are improved by combining a double-label system and intelligent optimization circulation, and the construction efficiency, the retrieval accuracy and the result traceability of the knowledge base are remarkably improved; the method is suitable for intelligent processing and application of long complex structure documents such as academic specialities, technical documents, policies and regulations and the like.

The invention relates to a semi-supervised remote sensing image classification method based on hierarchical crossing, and belongs to the technical field of remote sensing image processing and computer vision. The method comprises the following steps: firstly, constructing a representation learning module, adopting a hierarchical cross pseudo-tag generation mechanism, randomly combining different hierarchical features in a main model and an index moving average (EMA) model to construct a cross model, and generating a high-quality pseudo-tag for a weakly enhanced image; secondly, a self-adaptive weight mechanism is introduced, the label loss-free contribution is dynamically adjusted in combination with the training progress and the pseudo label utilization rate, and the self-adaptability of the model in different training stages is enhanced; and finally, designing a label alignment strategy based on a training stage, adjusting pseudo label category distribution through a time decline function, guiding the model to pay more attention to minority categories at the initial stage, and improving small sample category performance. Through hierarchical cross random combination, an adaptive weighting mechanism and a label alignment module, pseudo label quality, training stability and minority class performance are improved.

The invention discloses a commercial customer service system based on large model emotion recognition labeling and correction, and belongs to the technical field of natural language processing and deep learning. In order to solve the problems that an existing emotion recognition system depends on large-scale manual labeling, label quality is unstable and small sample performance is poor, a multi-model collaborative labeling and iterative optimization mechanism is adopted, and fusion labels are generated through automatic basic model selection, small sample LoRA fine adjustment, double-model divergence detection and large model arbitration. And a refining training set is constructed for iterative fine tuning to form a closed-loop optimization system. The method can effectively reduce the labeling cost, improves the label consistency and the emotion recognition precision under complex semantics, and is suitable for business information, financial public opinions and intelligent customer service scenes.

The invention relates to a semi-supervised small sample image classification method and system in a cross-domain scene, and relates to the technical field of deep learning and machine learning, and the system comprises a pre-training feature extractor and a self-supervised encoder; using a feature extractor and a self-supervised encoder to respectively obtain initial features and internal features, and fusing the initial features and the internal features to obtain fused features; generating initial pseudo labels through a clustering algorithm, and screening the initial pseudo labels according to a clustering validity index to obtain high-quality pseudo labels; constructing a mixed loss function, and updating the feature extractor on the mixed training sample by using the mixed loss function; and inputting a to-be-classified sample into the updated feature extractor, and outputting the category of the to-be-classified sample. According to the method, the technical problems of unstable model training and weak generalization ability caused by low pseudo label quality when small sample learning tasks with significant domain differences are processed in the prior art are solved.

The invention discloses an unsupervised domain adaptive medical image segmentation method based on multi-view alignment and pseudo label optimization, and aims to solve the problems of insufficient segmentation precision and low pseudo label quality caused by domain offset. According to the technical scheme, firstly, image level alignment is executed through a frequency domain smooth fusion module, and a class target domain image is generated; pre-training a segmentation network by using the image and generating an initial pseudo tag; then, through a two-stage optimization process, the integrity and the structural rationality of the pseudo tag are improved through prototype-based potential foreground completion and SAM-based structural perception enhancement in the process; and finally, on the basis of the optimized high-quality pseudo tag, constructing a multi-view prototype contrast learning framework to carry out final feature level alignment training. According to the method, the segmentation precision of the model on the label-free target domain is improved, and an effective scheme is provided for solving the challenge of scarcity of annotation data in medical image segmentation.

The invention discloses a coevolution driver cognitive load personalized quantification method, which relates to the technical field of traffic safety, and comprises the following steps: collecting a multi-modal physiological signal and processing the multi-modal physiological signal into structured feature data; constructing a labeled sample set D1 and an unlabeled sample set Du based on the structured feature data to perform semi-supervised collaborative pseudo labeling training, generating pseudo labels for unlabeled samples, and forming a self-labeled data set S; and constructing a training set and an enhanced training set by using S and D1, carrying out supervised contrast learning feature extraction, and carrying out model tuning in combination with an MAML algorithm. And finally, inputting any sample into the optimized model to generate continuous cognitive load value prediction. Wherein only a small amount of manual labeling is needed, the label quality can be improved, and the personalized quantification efficiency of the cognitive load can be improved; multi-modal information is fused, so that the robustness and the accuracy are improved; a continuous and fine-grained quantification mode is provided, and the accurate automatic driving decision-making requirement is met.

The invention discloses a flood storage and detention area change pattern spot identification label generation technology based on style decoupling. The technology comprises the following steps: S1, constructing and preprocessing a change detection data set; s2, constructing a conditional diffusion generation network based on a decoupling encoder; s3, decoupling extraction and orthogonalization representation of content-style features are carried out; s4, constructing a multi-target training strategy and two-stage model training; s5, injecting and fusing style features based on cross attention; s6, cross-domain style migration and diversified label generation; and S7, based on label quality screening of physical and semantic double constraints, outputting a high-quality change detection expansion data set. Compared with the prior art, the method has the advantages that by introducing a content-style decoupling mechanism, the style and the content of the generated sample are independently and accurately controlled, and the change detection label which is consistent in ground feature layout, diversified in imaging style and accurately labeled at a pixel level is generated.

The invention discloses a labeling quality intelligent detection method and system for multiple materials, and the method comprises the steps: S1, collecting original image data of paper, plastic and metal labels through an industrial camera, and carrying out the preprocessing, and obtaining standardized image data; s2, designing an improved edge detection network, extracting an initial edge feature of the label, generating a linear edge feature through curve fitting, calculating a label rotation angle, and executing rotation transformation to obtain image data after rotation correction; s3, constructing a double-branch network comprising material and defect branches, extracting material types, generating defect features in combination with the material types, and finally generating structured defect features; and S4, according to the structured defect features, generating material tolerance, and then calculating a quality score and a quality grade. According to the method, the problems of material misjudgment and defect detection omission caused by low rotation correction precision of the special-shaped label and staged processing of multi-material defect detection in a traditional method can be solved.

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 automatic visual detection, and discloses a bottle body label quality detection-based model automatic generation method, which comprises the following steps of: acquiring a multi-angle image of a bottle body, carrying out denoising, enhancement and brightness equalization processing, extracting a curved surface label through target detection and a perspective transformation algorithm, and carrying out image processing on the curved surface label; further generating a distortionless complete label image through an image registration and fusion technology, and taking the distortionless complete label image as a training sample set of a label defect detection model constructed based on a double-branch deep learning network; and performing hot updating on the label defect detection model through a closed-loop feedback and continuous learning mechanism. According to the method, the technical problems of imaging deformation of the curved surface label, diverse defects, difficulty in detection and performance degradation after model deployment are effectively solved, and high-precision, high-robustness and sustainable-evolution automatic label quality detection is realized.

The invention provides an industrial equipment state identification and health management method and system. The method comprises the following steps: acquiring comprehensive data of industrial equipment at each historical moment in a preset first historical time period; performing feature information calculation according to the operation data and the environment data at each historical moment; screening all the feature information to obtain target feature information; searching comprehensive data corresponding to the target feature information, and training a model based on the comprehensive data corresponding to the target feature information to obtain an industrial equipment state information identification model; and inputting the comprehensive data of the industrial equipment at the current moment into the industrial equipment state recognition model to obtain state information corresponding to the industrial equipment at the current moment, and generating a health management method according to the state information. According to the method, the key boundary samples are automatically mined through an intelligent screening mechanism, and meanwhile, a differential labeling strategy is adopted, so that the labeling cost is reduced while the label quality is ensured, and the final model recognition precision is also improved.

The embodiment of the invention relates to an annotation quality control method and device with active learning ability, and the method comprises the steps: constructing and training a difficulty evaluation model and a quality prediction model before executing a first document annotation task; in the labeling process of each piece, labeling personnel are dynamically distributed based on the evaluation result of the difficulty evaluation model and the personnel information base, the labeling quality is monitored in real time by using the quality prediction model in the labeling process, the labeling result is subjected to quality evaluation according to rules when labeling is finished, and the personnel information base and the labeling sample pool are updated; after the first document task is executed, whether active learning needs to be carried out or not is confirmed regularly according to the recent evaluation condition, if yes, typical records are selected from the labeled sample pool to update the verification data set, and the difficulty evaluation model and the quality prediction model are upgraded based on the updated verification data set. According to the invention, the efficiency can be improved, the cost can be reduced, real-time early warning and dynamic task allocation can be realized, and the quality control quality can be actively improved.

In various examples, data collection vehicles or machines may be equipped with one or more LiDAR sensors (and / or other sensors), and the LiDAR sensor(s) may be used to collect frames of LiDAR data representing various real-world conditions. The LiDAR data may be processed using one or more deep neural networks (DNNs) such as a transformer neural network to generate auto-labels representing detected dynamic obstacles of any designated class. Tracking may be applied to generate object tracks (tracklines), estimate velocity, and / or handle occlusions. In some embodiments, the object tracks may be refined based on geometry and / or confidence to improve their accuracy. In some embodiments, the auto-labels are classified to generate an estimated representation of quality, and auto-labels with at least a threshold quality score may be skipped during human labeling. As such, auto-label quality scores may be used to accelerate human validation of auto-labeled scenes by skipping high quality auto-labels.

The invention provides a self-adaptive target detection method, system and device in online testing, and relates to the technical field of computer vision and machine learning, and the method comprises the steps: obtaining an input image, and respectively inputting the input image to a fixed teacher model and a to-be-optimized student model to obtain a prediction result; calculating consistency distillation loss based on prediction results of the two parties to maintain the stability of the model, generating high-quality pseudo labels through a dynamic pseudo label generation process based on a teacher prediction result, and calculating supervision loss to drive the model to adapt to a target domain; and finally, combining the two losses to optimize the student model and outputting a detection result. According to the method, the technical problems of unstable pseudo tag quality and high model updating overhead in self-adaption during testing are effectively solved, and the precision, robustness and deployment efficiency of the target detection model in an unknown domain environment are remarkably improved.

The invention discloses a full-automatic label printed matter inspection machine, and belongs to the technical field of label quality inspection, the full-automatic label printed matter inspection machine comprises a rack, a workbench and an unwinding roller are respectively mounted below the front surface of the rack, a winding roller and a display screen are respectively arranged above the front surface of the rack, and guide rollers are rotatably connected to the left and right sides of the front surface of the rack at equal intervals; an industrial camera is installed on the front face of the rack and located below the display screen. The device further comprises a supporting plate, the supporting plate is installed on the front face of the workbench through bolts, a vertical lead screw is connected to the interior of the supporting plate through a bearing, and a first motor is installed at the bottom of the supporting plate. According to the full-automatic label presswork inspection machine, through cooperative driving of multiple motors and air cylinders and combination of an innovative structure of clamping inclined stripping, circulating stripping and magnetic force accurate pressing, full-process automation, zero-damage and high-thoroughness stripping of unqualified labels is achieved, the problems of residues and displacement are thoroughly solved, and the production efficiency and the stripping effect are greatly improved.

The invention provides a user portrait quality evaluation method and device, electronic equipment and a storage medium, and relates to the technical field of data processing, and the method comprises the steps: firstly, obtaining portrait tags of a user portrait in multiple dimensions; then, inputting the portrait labels of the multiple dimensions into a label quality evaluation model to obtain label quality evaluation results which are output by the label quality evaluation model and respectively correspond to the multiple dimensions; and finally, determining a quality evaluation result of the user portrait according to the label quality evaluation results corresponding to the multiple dimensions. In the embodiment, the label quality evaluation model is utilized to perform quality evaluation on the portrait labels of multiple dimensions, and then the quality evaluation result of the user portrait is determined, so that the one-sidedness of single-dimension evaluation can be avoided, and the accuracy of the quality evaluation result of the user portrait is improved.

The invention discloses an image information processing method, apparatus and device, and a computer medium. The method comprises the steps of obtaining multi-frame point cloud data of a vehicle surrounding environment collected by a laser radar; determining a three-dimensional point cloud map corresponding to the surrounding environment of the vehicle according to the multi-frame point cloud data; obtaining a look-around image collected by a camera; projecting the three-dimensional point cloud map to the look-around image to obtain a target image; and labeling the three-dimensional point cloud map based on the target image to obtain a labeled image, thereby playing a role in improving the accuracy of label generation and the label quality.

The invention discloses an unsupervised video clip retrieval method based on time sequence anchor point mining and semantic alignment, which comprises a video retrieval training system based on time sequence anchor point mining and point supervised learning, and the system comprises a key anchor point extraction module, a semantic alignment description generation module and a point supervised learning enhancement module. Key time anchor points are extracted from an original unlabeled video sequence through a key anchor point extraction module, then a pseudo-label triple is constructed through a semantic alignment description generation module, a weak supervision video clip retrieval model is trained accordingly, and in the training process, the pseudo-label triple is extracted from the original unlabeled video sequence. Constructing a point supervised contrast learning target through a point supervised learning enhancement module to further optimize the model, so that the finally trained weak supervised video clip retrieval model outputs a clip starting and ending time boundary related to query statement semantics; the problems of low pseudo tag quality, fuzzy positioning boundary and inconsistent semantic space in an unsupervised scene are effectively solved, and the positioning precision and generalization ability of video clip retrieval are remarkably improved.

The label quality inspection machine capable of automatically correcting and supplementing the materials comprises a first base and a second base, a screening device is fixedly installed in the middle of the right end of the first base, a first servo motor is fixedly installed on the front portion of the right end of the first base, and a controller is fixedly installed on the rear portion of the right end of the first base. A detection device is fixedly installed in the middle of the upper end of the second base, a conveying belt is installed between the first base and the second base, a feeding table is fixedly installed at the front end of the first base and the front end of the second base jointly, a second servo motor is fixedly installed on the left portion of the upper end of the feeding table, and a feeding roller is fixedly installed at the output end of the second servo motor. A first material receiving box is installed at the rear end of the first base and the rear end of the second base jointly. According to the label quality inspection machine capable of automatically correcting the errors and supplementing the materials, scanning detection is carried out through the visual inspection camera, and when the visual inspection camera detects that information of hang tags is fuzzy and is not correct, the unqualified hang tags are sucked up through the negative pressure suction cup.

The utility model discloses an auxiliary mechanism for labeling wires, which belongs to the field of wire processing and comprises a base, a first clamping component, a second clamping component and a driving component, the first clamping component comprises a first pneumatic finger, a first clamping piece and a second clamping piece, the first pneumatic finger is mounted on the base, and the second pneumatic finger is mounted on the second clamping piece. The first clamping piece and the second clamping piece are connected with the two output ends of the first pneumatic finger, the first pneumatic finger drives the first clamping piece and the second clamping piece to clamp or loosen a wire rod, the second clamping assembly is installed on the base, and the second clamping assembly is used for clamping or loosening the wire rod. The second clamping assembly and the first clamping assembly are arranged in a spaced mode in the preset direction, the driving assembly is connected with the first clamping assembly or the second clamping assembly, and the driving assembly is used for adjusting the distance between the first clamping assembly and the second clamping assembly in the preset direction. According to the utility model, the power line can be quickly positioned, clamped and straightened, the labeling quality can be improved to a great extent, and the consistency of the labeling quality is ensured.

The invention relates to a bridge-crossing type label-domain-free adaptive damage identification method, which belongs to the technical field of bridge structure health monitoring and comprises the following steps of: 1, constructing a three-tower type feature extractor comprising a time domain tower, a frequency domain tower and a modal tower to obtain a cross-modal combined feature; 2, aligning with a modal map through modal consistency constraint; 3, utilizing a multi-scale and dynamic frequency band attention mechanism to adaptively highlight a damage sensitive frequency band; 4, confrontation and cooperation mixed alignment is adopted, feature distribution differences are reduced, and modal physical consistency is kept; 5, a self-training and multi-teacher distillation mechanism guided by modal confidence is designed under the non-labeling condition, and the pseudo label quality and the training stability are improved; and 6, introducing physical constraint regularization to ensure that a result accords with a structural dynamics rule. According to the method, the damage probability distribution, the modal similarity heat map and the reliability index are finally output, target domain labeling is not needed, and the accuracy, the robustness and the engineering applicability of cross-bridge damage identification can be improved.

The invention relates to an image annotation method and device, a storage medium and electronic equipment. The method comprises the following steps: inputting to-be-labeled image data into a visual encoder of a preset multi-modal model for feature vector conversion processing to obtain an image embedding vector; obtaining a text set embedding vector corresponding to the preset text label set, the text set embedding vector being obtained by performing vector conversion on the preset text label set based on a text encoder of the preset multi-modal model, and storing the text set embedding vector as a model parameter of the preset multi-modal model; determining a target prediction result based on the image embedding vector, the text set embedding vector and a preset prediction model; and performing annotation processing on the to-be-annotated image data based on the target prediction result to obtain target annotated image data. Through the preset multi-modal model for storing the text set embedding vector, the data generation efficiency and the label quality are improved, the problem of identification range limitation caused by a label library generated based on image data is solved, open label expansion is supported, and the accuracy of image labeling is improved.

The application discloses a semi-supervised sea surface target detection method, system, device and storage medium, and relates to the field of target detection. A confidence threshold adjustment model of sea surface target categories is established, and a classification loss function of unlabeled data is determined according to the confidence threshold adjustment model; a regression loss function of the unlabeled data is adjusted through boundary box consistency regularization; and the confidence threshold is used to improve the pseudo-label quality in the model training process. The confidence threshold of each sea surface target category is dynamically adjusted, the pseudo-label quality is improved, the strong dependence on the intersection over union in the traditional positive and negative sample matching mechanism is relieved through boundary box consistency regularization, the quality of the boundary box is improved, and then the sea surface target detection performance is improved.

The application provides a low-bubble adhesive label quality monitoring method and system based on image analysis, and belongs to the technical field of adhesive labels. The low-bubble adhesive label quality monitoring method based on image analysis comprises the following steps: acquiring image acquisition data of a low-bubble adhesive label, wherein the image acquisition data is digital image data obtained by imaging the low-bubble adhesive label containing a pattern area under preset imaging conditions; performing image processing on the image acquisition data to generate processed image data for representing label region brightness distribution, gray scale distribution and texture distribution; analyzing image abnormalities formed by bubble points in an adhesive layer at corresponding apparent positions of a face material layer according to the processed image data to generate bubble point candidate region data; and performing quality determination on the low-bubble adhesive label according to the bubble point candidate region data to obtain a quality monitoring result.

The invention discloses a source-domain-free transfer learning target detection method based on double-adapter pseudo tag generation, and the method comprises the steps: firstly initializing a double-teacher model and a student model, and carrying out the data enhancement of a target domain image; introducing a confidence adapter, dynamically adjusting a confidence screening range through statistical analysis according to confidence distribution characteristics of each category, and generating candidate pseudo tags; introducing a category imbalance adapter, and fusing a confidence historical threshold and a current threshold according to the frequency of each category to obtain a current final confidence threshold; and finally, supervising student model training by using the screened pseudo labels, and updating double-teacher model parameters through index moving average to realize passive domain migration of the model. According to the source-domain-free target detection method based on double-adapter pseudo tag generation, the problems that pseudo tag confidence distribution is inconsistent and category samples are unbalanced are effectively relieved, and pseudo tag quality and cross-domain detection performance are remarkably improved.

The invention relates to a semi-supervised semantic segmentation method based on adaptive pseudo label generation, which realizes adaptive control of pseudo label quality by constructing category prototype representation containing a feature mean value and a standard deviation and introducing standard deviation information into a pseudo label screening and prototype consistency learning process. The method comprises a class prototype generation module, an adaptive pseudo-label generation module and a prototype consistency learning module, dynamically adjusts a pseudo-label screening threshold by combining prediction probability distribution and class prototype similarity, and introduces a prototype-based feature consistency constraint in a training process. Therefore, the segmentation precision and the training stability of the semantic segmentation model are remarkably improved under the condition of a small amount of annotated data.

The present application relates to a kind of heart ultrasound video semantic segmentation pseudo-label generation method based on optical flow guide, belong to computer vision and medical image processing field.The method includes: selecting two key frames in heart ultrasound video frame sequence, and obtaining the segmentation mask of two key frames by artificial labeling;Optical flow model is fine-tuned using two key frames, the frame between two key frames, the frame of first key frame left side preset quantity and the frame of second key frame right side preset quantity;Based on the fine-tuned optical flow model, the forward optical flow sequence from first key frame to second key frame and the reverse optical flow sequence from second key frame to first key frame are predicted;Based on forward optical flow sequence and reverse optical flow sequence, generate forward propagation mask sequence and reverse propagation mask sequence, and carry out position weighted fusion, obtain the pseudo-label of unlabelled frame.The technical problem that the present application aims to solve is that the label of heart ultrasound video semantic segmentation is difficult to obtain and the quality of pseudo-label obtained is poor.

The present disclosure relates to a solution on label quality indicator for model positioning based on PRU is proposed. In particular, a target UE receives a signal from LMF to consider the assessment of the label quality using the PRU information based on the target UE capability. Furthermore, the target UE reports the quality indicator of the label. In this way, the accuracy assessment of the label used for data collection for AIML positioning can be improved.

The application provides a labeled sample data generation method and an electronic device, and the method comprises the following steps: acquiring multiple frames of images for a target scene; determining the marking position information of each marking object in each frame of image, and determining the initial spatial pose of each marking object in each frame of image; determining the target coordinate system corresponding to the multiple frames of images; determining the global spatial structure information corresponding to the multiple frames of images according to the multiple frames of images, the target coordinate system corresponding to the multiple frames of images, the marking position information of each marking object in each frame of image and the initial spatial pose of each marking object in each frame of image; and generating labeled sample data based on the global spatial structure information corresponding to the multiple frames of images and the labeling information of the target object in the multiple frames of images. The application can realize the collaborative improvement of multi-view key point labeling in efficiency, multi-view consistency and controllability of labeling quality without relying on complex external three-dimensional reconstruction process or special calibration software.

The utility model discloses a self-adhesive label quality detection device, which relates to the field of self-adhesive label production, and adopts the technical scheme that the self-adhesive label quality detection device comprises two symmetrical half shells, detection mechanisms are mounted in the half shells, through grooves for the self-adhesive labels to penetrate through are formed in the two sides of the half shells, and the detection mechanisms are arranged in the through grooves. During use, the two half shells are lifted and then attached to each other to be locked after being lifted to proper positions, so that the upper edge and the lower edge of a self-adhesive label are covered when the self-adhesive label is driven by the winding device, the self-adhesive label penetrates through the complete shell to move in the through groove, and then the self-adhesive label is wound into the through groove. The quality of the upper surface and the lower surface of the self-adhesive label is detected through the detection mechanisms on the upper side and the lower side, so that the double-side detection of the self-adhesive label can be directly completed in the moving process of the self-adhesive label, the working procedure steps and time for producing the self-adhesive label are reduced, meanwhile, the self-adhesive label does not need to be additionally moved, and the occupied space is reduced.