48 results about "Multi-label classification" patented technology

This application proposes a problem classification method, a problem classification device, and a computer storage medium. The problem classification method includes: acquiring mathematical problem information; extracting mathematical features from the mathematical problem information, wherein the mathematical features include text features, formula features, and / or graphic features; performing logical reasoning on the mathematical features according to a preset mathematical logic chain; inputting the logical reasoning result into a multi-task output branch to determine the problem classification result of the mathematical problem information. This problem classification method achieves end-to-end multi-label classification of mathematical problems.

This invention belongs to the technical field of natural language processing and dialogue systems. It discloses a multi-target dialogue recommendation method based on hierarchical prompt optimization, including hierarchical target label modeling, learnable prompt embedding initialization, graph attention enhancement mechanism, multi-task learning framework, and response generation model. Hierarchical target label modeling encodes the hierarchical relationships between target labels into a graph structure. Learnable prompt embedding initialization integrates label semantics and deep hierarchical information into the encoder input space. The graph attention enhancement mechanism enhances the label vector structure perception ability through information transmission and aggregation between label prompts. The multi-task learning framework jointly optimizes masked language modeling and hierarchical multi-label classification. The response generation model uses the predicted target as a control signal to guide generation. This method effectively models the hierarchical dependency relationship between target types and target entities, enhances the structure perception ability of label vectors, maintains the model's language understanding ability, and improves the accuracy of target prediction.

This invention discloses a method and system for identifying apparent damage to bridge concrete based on texture analysis, belonging to the interdisciplinary field of bridge engineering inspection and computer vision technology. The method includes: acquiring surface images of bridge components using a drone and generating orthophotos; extracting contrast, correlation, and entropy features based on the gray-level co-occurrence matrix; extracting histogram features based on local binary patterns; detecting corrosion and seepage areas based on the HSV color space; fusing multi-dimensional features and inputting them into a multi-label classifier, simultaneously outputting pixel-level segmentation masks for multiple types of damage such as cracks, spalling, exposed rebar corrosion, and seepage; generating a damage level score and maintenance recommendation report according to bridge technical condition assessment standards. This invention integrates complementary texture features to achieve simultaneous identification of multiple types of damage, outputs pixel-level segmentation results, and automatically generates an assessment report that conforms to engineering specifications.

In various examples, multilabel hierarchical classification of objects for autonomous systems and applications is described herein. Systems and methods are disclosed that use one or more neural networks to classify objects, such as traffic signs, using multilabel classification and / or hierarchical classification. For instance, a multilabel subnetwork of the neural network(s) may classify an object based at least on one or more attributes associated with the object. As such, the output from the multilabel subnetwork may include at least a classification associated with the object and an attribute classification(s) associated with the object. A hierarchical subnetwork of the neural network(s) may also classify the object using one or more class labels, where a class label indicates another classification and / or a class group associated with the object. The systems and methods may then use the classification, the attribute classification(s), and / or the class label(s) to determine a final classification associated with the object.

The present application relates to the technical field of medical image processing, in particular to an ANA cell image multi-mode analysis and prediction method, system, device and medium, through a target detection network, interval cells in an original image are detected, a dynamic sampling mechanism is triggered to cut and sample a sliding window of continuous multiple frames of effective images, and a sample image block set obtained by sampling is input into a constructed multi-label classification network, a CBAM attention module and a category decoupling attention module are embedded in a backbone network of the multi-label classification network, enhanced features are extracted, and the probability that a detection image belongs to one or more categories of cell nucleus types is output. The present application reduces background interference through a dynamic sampling mechanism, enhances small sample category recognition ability through category decoupling attention, optimizes multi-label output through category correlation punishment, and improves the accuracy and clinical practicability of ANA cell image multi-mode analysis.

The present disclosure relates to a multi-label classification model training method and device, and a storage medium. The method comprises: dividing a single-label picture dataset into multiple parts to obtain multiple original sub-training sets; sequentially performing the following processing on each original sub-training set: performing class inference on the current original sub-training set according to the current multi-label classification model to obtain the inference result of each picture in the current original sub-training set, and selecting at least one picture to be labeled from the current original sub-training set according to the label category and the inference result of each picture, wherein the at least one picture to be labeled meets a to-be-labeled condition; obtaining at least one labeled picture according to the picture to be labeled, and forming a target sub-training set by using the at least one labeled picture; and training the current multi-label classification model according to the target sub-training set. The technical solution selects a small amount of pictures for labeling, maximizes the training effect of labeling all batch data, thereby saving labeling costs and improving model training efficiency.

The application discloses a kind of based on defect feature's weld bottom plate image identification method and system, it is related to ray nondestructive testing technical field.The method is first obtained weld bottom plate image and is preprocessed, then accurately extract weld area as region of interest, and from the area inside parallel extraction morphological geometry, texture and gray scale statistics etc. Multi-scale feature, and carry out intelligent fusion based on dynamic weighting, generate fusion feature vector;While extracting the global attribute feature of image, and input pre-trained feature utility prediction model, the model is according to image quality adaptive output optimal feature subset selection decision, according to this from fusion feature vector filters out optimal feature subset, then the feature subset is input multi-label classifier, to identify multiple defect types simultaneously, finally for the defect identified, according to digital rating standard automatically mapping output quality rating result, whereby can improve the accuracy, robustness and automation degree of weld defect identification.

This invention belongs to the field of multimedia analysis technology and discloses a multimodal media tampering detection method, system, device, and medium based on multi-view contrastive learning. The method includes: acquiring a training dataset, which includes training image-training text pairs and corresponding tampering category labels; introducing a cross-encoder based on a visual-language model, setting several multilayer perceptron head structures, and designing three contrastive learning methods: noise enhancement, prototype-based, and multi-label tampering classification, to obtain an initial multi-view contrastive learning framework; training the initial multi-view contrastive learning framework based on the training dataset to obtain a trained multi-view contrastive learning framework; and performing a tampering detection task on the image-text pair data to be detected based on the trained multi-view contrastive learning framework. The technical solution of this invention can improve the accuracy and robustness of multi-label classification.

This invention discloses an underwater video multi-label classification processing method and system, relating to the field of video processing technology. The method includes: processing underwater video to obtain original image features, clear image features, and text semantic features; encoding based on a multi-head attention encoder, fusing the original visual feature representation and text feature representation through a cross-modal attention mechanism to generate a multi-modal complete feature representation; using a generative adversarial network, utilizing the multi-modal complete feature representation, original visual feature representation, clear visual feature representation, and text feature representation as self-supervised signals to generate enhanced and completed multi-modal features; constructing a composite objective function, and performing multi-label classification of the underwater video based on the optimized model and the enhanced and completed multi-modal features. The multi-head attention encoder analyzes and processes the original video data, and constructs a self-supervised signal based on clear visual features, improving the accuracy of underwater video feature fusion.

The application discloses an event multi-label classification method and device based on a knowledge graph, equipment and a medium. In the application, text data is acquired, various entities, attributes and attribute values of the various entities contained in the text data are determined, and attribute triplets and relationship triplets are determined. The attribute triplets and the relationship triplets can contain more potential semantic information in the text data. Therefore, according to the attribute triplets and the relationship triplets, an event knowledge graph corresponding to the text data is constructed, and an event category corresponding to the text data is determined based on an event classification model, so that more semantic information can be mined to realize accurate event classification. Compared with a technical solution for event classification based on a word vector, a keyword and other technical means, the accuracy of event classification is improved.

This invention provides a method and system for automatically generating forensic reports, relating to the field of judicial information technology. The method includes: acquiring multimodal forensic materials of the object to be identified; extracting key information tags using a multi-label classification model to generate a tag set; constructing a prompt template for forensic identification based on the tag set using an automatic prompt search algorithm; determining key discovery description paragraphs by combining the prompt template and a language model; performing slot processing on the key discovery description paragraphs and the standard forensic report template using a slot classifier to determine the target position of the template; filling the key discovery description paragraphs into the target position of the template and optimizing the description content in the filled template to generate a draft forensic report; determining whether the draft forensic report passes quality verification; if so, outputting the draft forensic report as the forensic report; otherwise, adjusting the key discovery description paragraphs according to the quality verification results.

Embodiments of the present specification disclose a text classification method and device, a storage medium and an electronic device. After a first label set corresponding to a text to be classified is recalled, a second label set of the text to be classified is obtained through a preset co-occurrence probability matrix. In a selection stage for a target label corresponding to the text to be classified, a multi-label classification problem is converted into a matching problem between a label and the text to be classified. The first label set recalled and the second label set obtained through the co-occurrence probability matrix are matched with the text to be classified for binary identification, so as to finally determine the target label corresponding to the text to be classified.

The application discloses a kind of honeycomb sandwich composite structure ultrasonic phased array detection and honeycomb classification intelligent evaluation method, equipment, medium and product, it is related to non-destructive testing field, the method comprises: using ultrasonic phased array probe to collect the data of honeycomb sandwich composite structure to be detected, and based on glue layer and honeycomb interface pulse reflection wave carries out ultrasonic C scanning imaging to obtain ultrasonic image;Based on U-Net architecture, combined with the hybrid encoder-decoder structure of depth residual network and Transformer model is constructed multi-label classification network, to carry out multi-class probability prediction to each pixel point in ultrasonic image, and obtain probability prediction result;Based on probability prediction result and confidence threshold, the classification result of different bonding state is obtained;The classification result of different bonding state is visualized, and classification reconstruction map is obtained.The application can realize pixel-level multi-label semantic segmentation, and then realize accurate detection and classification of honeycomb in true sense.

The application discloses a kind of based on grouping consensus's incomplete multi-view multi-label classification method and system.Firstly, the potential distribution characteristics of heterogeneous data are extracted by independent encoder;Secondly, a dynamic soft route is constructed by adding semantic anchor points in the latent space to adaptively calculate the weight of each semantic subgroup to which the view belongs;Then, the weight is used to perform intra-group weighted expert product aggregation to generate multiple denoised grouping consensus features;Finally, these grouping consensus features and label prototype vectors are spliced in the feature dimension, and the deep interaction and prediction of heterogeneous semantics and label prior are realized through a unified classification head;At the same time, a two-stage training strategy is used to solve the routing cold start problem.The application effectively overcomes the view heterogeneity and noise interference in incomplete data, significantly improves the accuracy and robustness of multi-label classification.

The application discloses a kind of social governance risk element discovery method and system based on knowledge guide counterfactual reasoning, belong to artificial intelligence and public governance cross technical field, wherein method includes: based on historical case information construction knowledge graph, using semantic encoder extracts semantic features from historical case information, using graph neural network embedding in the node in knowledge graph, obtain knowledge characteristics;In the concept cluster level of knowledge graph, semantic level counterfactual intervention is executed, and fact fusion and counterfactual fusion are carried out;By the difference between fact fusion feature and counterfactual fusion feature, the fact fusion feature is corrected, and the cause-effect enhanced feature is obtained;Cause-effect enhanced feature is input into multi-label classification head, and risk element list is predicted.The application effectively improves the logical rigor, explainability and generalization ability of risk element identification by deeply integrating structured governance knowledge and causal reasoning without a large number of labeled causal labels.

The application provides an image classification method and device, equipment and a storage medium. The method comprises the following steps: obtaining a first training set, the first training set comprising a plurality of images, each image being labeled with a class label corresponding to each image category; constructing a structure of a first neural network model for multi-label classification; training the constructed first neural network model according to the first training set to obtain a multi-category classification model. The application realizes the simultaneous classification of multiple image categories through one model. The model learns the correlation between the features of different categories through a bidirectional long short-term memory network, and the learned features have better robustness. If the classification accuracy of a certain image category is relatively low, a branch model corresponding to the image category is also trained, the model is corrected by using the branch model, and the problem of low classification accuracy of individual image categories is solved. The model is trained by combining a cross-entropy loss function and a smoothing loss function, the training is more stable, and the effect is better.

The application discloses a multi-label classification method and system for a ticket assigning scene, and the method comprises the following steps: performing feature extraction on input ticket text to generate global text semantic representation; based on pre-constructed semantic embedding vectors of each preset business label, performing label knowledge enhancement on the global text semantic representation to realize alignment of text representation and label semantics in a semantic space; based on the aligned text semantic representation, calculating original confidence corresponding to each preset business label by a main classifier; for each preset business label and ticket text, dynamically generating a classification threshold corresponding to the business label; based on the classification threshold, judging the original confidence to output a multi-label classification result corresponding to the ticket text, so as to realize a differentiated ticket assigning decision. Therefore, by means of the dynamically generated individualized classification threshold, the misassignment and missed assignment problems caused by a fixed threshold in the ticket assigning scene are effectively solved, and the accuracy of multi-label classification is significantly improved.

This application provides a multi-label classification method for drainage pipe networks. It employs a convolutional neural network to extract multi-level image features from sewer images; and a graph convolutional neural network to extract defect label information from the sewer images. The label information is then used as a query, and the multi-level image features are fused as keys and values ​​into a Transformer decoder. Finally, the cross-interest module in the Transformer decoder is used to query whether relevant features exist in the sewer images, resulting in a multi-label defect classification of the sewer images. This method dynamically aligns image content with defect semantics through cross-modal interaction, thereby achieving a more accurate multi-label defect classification of sewer images.

The application discloses a rice germplasm yield prediction method and system for panicle type classification and growth period alignment. Firstly, time sequence images of a rice plot are acquired, and a ST-YOLO single panicle detection model is used to identify the growth stage of the rice and extract single panicle images. The single panicle images are input into a PP-LCNet multi-label classification model to synchronously identify the panicle type and the growth period, filter mature period data, and process the data by panicle type, so that the growth period alignment and the panicle type classification are realized, and the temporal and spatial feature noise introduced due to different phenology and morphology is eliminated. The geometric features of the detection frame are extracted at the single panicle scale, and the number, color, texture, spectrum, phenology and other features are extracted by using a Swin Transformer segmentation model at the group scale. Machine learning algorithms are used to respectively establish yield prediction models for different panicle types. The application effectively solves the problem of complex temporal and spatial noise interference in large-scale breeding tests, and realizes high-throughput, accurate and non-destructive rice yield prediction.

The application discloses a power distribution terminal equipment fault intelligent diagnosis method and system based on an Internet of Things, which comprises the following steps: S1, collecting multiple operation data of a power distribution terminal to construct a time sequence sample; S2, extracting power disturbance features based on a sliding window, generating a behavior coupling matrix and a topological connection matrix, and calculating node redundancy; S3, fusing two types of relations to construct a dynamic atlas; S4, inputting a graph diffusion network, fusing structure and behavior diffusion results, and generating a node state feature vector; S5, inputting a multi-label classification model to identify fault types and probabilities; S6, combining fault types and redundancy to generate a response instruction; S7, executing the response and updating a relation matrix and a classification model, and closing loop optimization of a diagnosis process. The application realizes accurate identification and rapid response of power distribution terminal faults, and improves intelligent operation and maintenance level of a power supply system.

A sequential feedback prediction system predicts a risk of an event associated with a set of potential outcomes. The system receives training data including data records that have information associated with risk factors. Each data record may be labeled with outcomes that have inherent sequential dependencies. The system trains sub-models, each sub-model predicting risks for a respective outcome using a set of risk factors. For each sub-model, the system may determine a subset of sub-models whose results are used as input for the sub-model. The system determines an order to run the set of sub-models such that prediction results for preceding sub-models may be used as inputs for subsequent sub-models. The system may determine a number of rounds to run the ordered sequence of sub-models until the performance of the ensemble model meets a predetermined threshold. The system may generate a risk score for each of the set of outcomes.

The application discloses a fuzzy multi-label inference learning method for heterogeneous data, comprising: label-aware enhanced feature construction, selecting a neighbor based on a hybrid difference metric and constructing a label-aware enhanced feature vector according to a difference relationship of positive and negative neighbor weight cumulative values; fusion enhanced antecedent construction, splicing the original feature and the enhanced feature and then obtaining a fuzzy feature vector through fuzzy rule activation; label structure guided consequent alignment, constructing a label difference matrix based on a statistical correlation coefficient between labels and taking the label difference matrix as a structured alignment constraint term acting on a consequent parameter matrix; joint optimization and prediction, integrating a data fitting term and the structured alignment constraint term into a unified objective function for optimization and solution to generate a multi-label prediction result. The application solves the problems of feature mapping modeling difficulty, insufficient learning of scarce labels and label dependency relationship interference in the heterogeneous data scene, and improves the accuracy and stability of multi-label classification.

This application discloses a multi-label classification method and related apparatus. The method includes: inputting data to be classified into a multi-label classification model; outputting a target category prediction result based on the data to be classified through a category prediction layer in the multi-label classification model; and outputting a target mutual exclusion group prediction result based on the target category prediction result through a mutual exclusion group prediction layer in the multi-label classification model; correcting the target category prediction result based on the mutual exclusion group prediction result to obtain the target multi-label classification result corresponding to the data to be classified. This method not only enables the mutual exclusion group prediction to more accurately capture the logical relationship between label categories, avoiding misjudgment of mutual exclusion relationships caused by independently using the original features of the data to be classified, thus improving the reliability of model prediction, but also effectively eliminates logical conflicts in the target category prediction results, avoiding the contradiction of simultaneous occurrence of mutually exclusive label categories, and obtaining a target multi-label classification result that better conforms to actual business logic, thereby improving the accuracy of multi-label classification.

This invention relates to a method for identifying bridge rubber bearing defects based on target detection and multi-label classification, comprising: acquiring original images of bridge rubber bearings; extracting rubber bearings from the original images using a trained target detection network to obtain multiple prediction boxes; extracting rubber bearing images corresponding to each prediction box from the original images based on the prediction boxes; classifying the rubber bearing images using a trained defect classification network; and coupling the overall state vector and defect label vector to obtain the final identification result of the rubber bearings. Through a cascaded framework of "target detection-defect classification," defect identification is decoupled into two stages: localization and classification. This effectively overcomes interference from small targets and background, and innovatively proposes a two-layer classification network of "overall state + defect label," ensuring the mutual exclusivity of "intact / uncertain / defect" conclusions, completely avoiding contradictory outputs, and significantly improving identification accuracy, logical rigor, and engineering practicality.

This invention discloses a method for predicting the function of bioactive peptides based on multi-view, multimodal representation learning. The method includes: extracting amino acid sequence information of peptides using multi-scale dilated convolutional CNN and bidirectional LSTM; extracting structural and functional features of peptide molecules using an ESM-2 model; processing molecular fingerprint information using convolutional CNN and Mamba structures; extracting topological information of the peptide molecular graph using traditional convolutional CNN, and processing node features using graph convolutional neural networks. All these multi-view features are ultimately fused into an aggregated feature representation, which is then passed through a fully connected layer and a sigmoid function is applied for multi-label classification. By concatenating and fusing the extracted features, a comprehensive peptide molecule feature representation is formed to predict various bioactive properties of the peptide. This multi-view, multimodal feature integration method not only enhances the model's predictive ability but also improves its flexibility and accuracy when processing complex biological data.