125 results about "Class information" patented technology

The invention provides a drug temperature and humidity digital monitoring and quality tracing system based on a full life cycle, and relates to the technical field of digital monitoring, and the system comprises the steps: firstly collecting the temperature and humidity data of the whole link of drug production, storage, transportation to terminal sales, and obtaining a standardized data set through the verification and packaging of an industrial standard library; performing secondary screening through primary screening in combination with a dynamic temperature and humidity screening model to obtain a precise abnormal data set; associating drug full-cycle multi-class information, and constructing a temperature and humidity-drug state mapping relation based on a time sequence association algorithm to form a full-dimension association data set; based on the data set, combining with a quality risk assessment model for early warning, realizing full-period traceability through a reverse traceability algorithm, and generating an assessment result and a traceability link; and finally, relevant data are converted into a dynamic temperature and humidity monitoring curve, a quality risk early warning report and a full-life-cycle tracing atlas, and an interactive report is assembled, so that the temperature and humidity state of the medicine can be checked conveniently.

The invention discloses a low-voltage uninterruptible operation site visual monitoring optimization method based on an edge computing architecture, and the method comprises the following steps: constructing an operation scene graph analysis module, obtaining the distribution of operators, the operation type and the equipment voltage class information of each monitoring region, generating a risk class mapping graph, and carrying out the operation scene graph analysis module; positioning and initial identification of a high-risk area are realized; generating a risk perception matrix; constructing a priority queue of computing resources and risk levels; acquiring identification precision, processing delay and behavior capture integrity indexes, and establishing a task feedback mechanism; constructing a scheduling logic credibility evaluation module, analyzing a scheduling instruction, marking and recording an abnormal path, generating a scheduling behavior record library, and predicting and early warning a priority inversion risk; and iteratively optimizing scheduling model parameters, and dynamically updating risk scores and scheduling logic. According to the method, high-risk task resources are guaranteed preferentially, priority reversal is avoided, and the stability, robustness and safety response timeliness of the system in a complex operation environment are remarkably improved.

The invention provides a multi-dimension-based information identification method and device and electronic equipment, and relates to the technical field of data processing, and the method comprises the steps: carrying out the data analysis processing of a plurality of pieces of information data transmitted by an identified risk transmitting end, determining a receiving end, and obtaining the information interaction data between the risk transmitting end and the receiving end; performing classification processing on the multiple pieces of information data according to information interaction data to obtain one-way information data and interaction information data; inputting the one-way information data, and / or the interaction information data and the interaction information into a trained information identification model for information identification processing to obtain an information probability value; and determining the information data with the information probability value greater than a preset probability threshold as abnormal information data. The condition that a single piece of information is normal but can be determined as abnormal information after multiple pieces of information are integrated can be effectively identified, so that the accuracy and comprehensiveness of information identification are improved.

A computer acquires training data including first text, first class information indicating a class mapped to a single word contained in the first text, first position information indicating a position of the single word in the first text, and first range information indicating a range of a first named entity that includes the single word in the first text. The computer executes, based on the training data, machine learning of a machine learning model which is used to estimate, from text, class information, and position information, range information of a named entity included in the text.

The invention provides a camouflage object semantic segmentation method and device based on an adaptive candidate strategy, and a medium, and the technical scheme of the invention obtains the category code of a to-be-segmented image through a coarse feature extractor and a classifier, achieves the transmission of semantic category information, and enhances the semantic perception capability of a subsequent segmentation task. Initial prediction of a camouflage object is carried out through a target detector and an adaptive candidate strategy, an optimal target attention box is generated, and robust space guidance is provided for a core segmentation task; according to the method, category coding and a target attention box are combined, multi-source information is fused and reconstructed through a multi-guide feature fusion module and a multi-task perception decoder, and an accurate segmentation result is output.

An example operation may include at least one of producing, by a class-conditioned sample generator executing on at least one processor communicatively coupled to a memory on a host platform, a synthetic feature set based on a label sequence and class information derived from received data, transmitting, by the host platform, a finalized synthetic sample to a computing device when the synthetic feature set satisfies a fidelity threshold, generating, by the computing device, a fidelity score based on a comparison of the finalized synthetic sample to the label sequence and the class information, retraining, by the computing device, the class-conditioned sample generator based on the fidelity score, and validating, by the computing device, the class-conditioned sample generator by transmitting a test prompt to the host platform, receiving a synthetic response generated by the class-conditioned sample generator, and comparing the synthetic response to previously stored synthetic data to validate the class-conditioned sample generator.

The invention discloses a warehouse management method and system based on the Internet of Things, and the method comprises the steps: determining a target material and the demand data of the target material according to a production task issued by a manufacturing execution system; obtaining attribute information of the target material based on the demand data, and distributing priority for the target material according to the attribute information to obtain priority information of the target material; wherein when the target materials comprise various materials, the priorities of different types of materials in the target materials are distributed according to first-class information, and the priorities of the same type of materials in the target materials are distributed according to second-class information; obtaining layout information of a current platform, wherein the layout information comprises storage position information and transportation path information; and according to the priority information of the target materials, the storage positions of the target materials on the current platform are adjusted, so that the target materials with higher priorities can be delivered out of a warehouse in a shorter transportation path. The storage position and the transportation path of the material can be adjusted according to the priority information of the material, and effective operation of a production link is ensured.

The invention discloses an electromechanical system heterogeneous domain generalization fault diagnosis method and device based on digital twinning, and belongs to the technical field of electric data processing. The method comprises the following steps: constructing a digital twin model to simulate entity characteristics of an electromechanical system, generating virtual fault data and structured interconnection knowledge distributed across working conditions through the digital twin model, and reconstructing a K-class information domain according to the virtual fault data and entity data of the electromechanical system; the information source is divided into a source domain and a target domain, the source domain represents visible operation conditions, and the target domain represents invisible operation conditions; performing structured interconnection knowledge embedding on the source domains to obtain a knowledge embedding data set of the K-class source domains; and constructing a heterogeneous domain generalization network, and training the heterogeneous domain generalization network by utilizing the knowledge embedding data set of the K-type source domain and the K-type target domain data set to obtain a fault diagnosis model so as to perform fault diagnosis on invisible operation condition information to be diagnosed. According to the invention, fault diagnosis can be carried out on the electromechanical system under the limitation of invisible operation conditions.

The invention discloses a mixed-flow water turbine runner surface cavitation point detection method. The method comprises the following steps: acquiring a mixed-flow water turbine runner surface cavitation point image to be detected and a trained cavitation point detection model; the method comprises the following steps: scaling a to-be-detected turbine surface cavitation point image of a mixed-flow water turbine to a preset specification, then sending the image to a backbone network, performing shallow feature extraction through a convbn module in the backbone network, and then performing deep feature extraction through a uib module so as to gradually extract image features of each level through the backbone network; inputting the image features of each level into a Neck network in which a BiFPN module is introduced and a high-resolution detection layer is added, and performing cross-scale feature interaction fusion on the image features of each level through the Neck network to obtain a multi-scale to-be-detected cavitation point feature map; the multi-scale to-be-detected cavitation point feature map is sent into a detection head, the positions and confidence coefficients of cavitation points in the map are predicted, accurate positioning and classification of cavitation points of different sizes are achieved through a dynamic self-adaption mechanism of the detection head, and a detection result containing bounding box coordinates, confidence coefficient scores and category information is output. According to the invention, water turbine runner surface cavitation point detection under different scales can be realized.

Use of behavior graphs is disclosed. Information is acquired related to datacenter activity comprising entry point information associated with a client entering a datacenter from an external entry point, a user on a machine class information, information on launched processes, child processes, and / or interactive processes, and information related to addresses with which processes communicate. Various tiers of nodes are generated based on the acquired information. A baseline graph is generated and used for comparison with subsequent behavior graphs. Selective remediation can be performed.

The invention provides a target detection method and device based on a meta universe, electronic equipment and a storage medium, and the method comprises the steps: obtaining an initial image, carrying out the AI creation of the initial image in the meta universe to obtain an enhanced image with a label, training a preset teacher model according to the enhanced image and the initial image, and obtaining a target detection result; and optimizing the weight of the teacher model through a gradient descent algorithm to obtain a trained target teacher model and a teacher model weight, performing weight updating on a preset student model through an index moving average and the teacher model weight to obtain an initial student model, inputting the initial image to the initial student model, and obtaining the target teacher model. Obtaining a prediction bounding box and prediction category information, calculating distillation loss based on the prediction bounding box and the prediction category information, performing knowledge distillation on the initial student model according to the distillation loss to obtain a target detection model, and performing target detection through the target detection model; the target detection method based on the meta universe is more practical and more accurate.

The present application relates to the technical field of medical image recognition, in particular to a semi-supervised medical image classification method and system based on similarity threshold pseudo label, more unlabeled data is used by using similarity threshold pseudo label, the representative probability distribution of each class is calculated according to labeled data, and labeled data can provide more reliable class information;Unlabeled data is selected according to the confidence threshold to give pseudo label, then the probability distribution of the low confidence part of unlabeled data and the similarity between the representative probability distribution of the corresponding class is greater than the similarity threshold, and the unlabeled data is given pseudo label, so as to select more high-quality unlabeled data to participate in model training.The present application can utilize more high-quality unlabeled data on the confidence threshold method without reducing the accuracy of pseudo label, and further improve the robustness of model training and the accuracy of medical image classification.

The disclosure provides a camouflage object semantic segmentation method and device based on an adaptive candidate strategy, and a medium. The technical scheme of the disclosure obtains the class code of a to-be-segmented image through a coarse feature extractor and a classifier, realizes the transmission of semantic class information, and enhances the semantic perception ability of a subsequent segmentation task. An initial prediction of a camouflage object is performed through a target detector and an adaptive candidate strategy, and an optimal target attention box is generated to provide robust spatial guidance for a core segmentation task. The class code and the target attention box are combined, multi-source information is fused and reconstructed through a multi-guided feature fusion module and a multi-task perception decoder, and an accurate segmentation result is output.

To implement quick and smooth parking assistance close to a human sense without being affected by an environment of a parking space. A 3D semantic segmentation image having a plurality of pixel units, each pixel unit of the plurality of pixel units including the depth data and class information, are generated, an available parking space is searched for on the basis of the 3D semantic segmentation image, and a path to the searched parking space is planned and a vehicle is controlled. The present disclosure can be applied to a parking assistance system.

The invention discloses a double-classification head voiceprint recognition training method based on federal learning. The process of the training method comprises the steps that a client provides local speaker category information, a server integrates global speaker categories, the server initializes a global classifier and a global encoder, the server distributes the global classifier and the global encoder, and the client receives the global classifier and the global encoder and then carries out training together with a local classifier. In the local training process, global classifier parameters are updated through a moving average algorithm, cooperative training of the global classifier and a client local classifier is provided, and the model performance is remarkably improved. In the model fusion process, in order to solve the influence caused by data distribution deviation, a global classifier carries out weighted averaging according to categories contained in each client. According to the method, the local model with better performance can be obtained, so that the global model performance loss during model aggregation is reduced, and a forward feedback mechanism is formed for subsequent model training.

The application discloses a kind of edge cloud cooperation human activity recognition modeling methods based on heterogeneous multi-modal data, and is divided into two stages of centralized pre-training and multi-modal semi-supervised fine-tuning.In the first stage, by dynamic mask contrast learning, pre-train the basic model based on converter using the joint dataset with super-class information, so that it has the ability to extract robust features from any modal combination.In the second stage, the local unlabeled data is used by the end-side client to generate pseudo-labels through weak mask view and calculate losses based on strong mask view for local update;Cloud side aggregates heterogeneous model parameters from different clients, and fine-tunes the global model using a small amount of labeled data.The application effectively alleviates the data heterogeneity problem by separating single-modal feature encoding and cross-modal information fusion, and can efficiently train a high-performance multi-modal human activity recognition model using a small amount of labeled data and unlabeled data while protecting data privacy.

The invention relates to an underwater vehicle intelligent navigation large model pre-training method and system and a medium. The method comprises the following steps: M1, multi-source data collection and format unification: collecting control of an unmanned underwater vehicle, a remote control underwater vehicle and a manned underwater vehicle in a normal navigation state, control in an abnormal navigation state, multi-device cooperative navigation control and control in different navigation environments; data information of control under different navigation depths and control in different navigation time periods is acquired, and preprocessing is carried out; m2, constructing a pre-training task: dividing sample points at each moment into state information and control information according to an execution process; according to the method, the generalization performance and the multi-task processing capacity of the model can be remarkably enhanced, more reliable technical support is provided for UV autonomous control, the problem of error transmission in a traditional multi-stage model assembly line is effectively solved, the maintenance cost is reduced, and meanwhile the robustness of the model is improved.

The present application relates to a kind of unmanned aerial vehicle micro target identification inspection method and system, belong to computer vision technical field.The present application utilizes the powerful computing power of cloud computing, based on DETR original model constructs micro target detection model, micro target detection model utilizes the query guide feature sequence generated by dynamic anchor frame module to initialize the position query of decoder, simultaneously, the present application will static learnable query vector as the input of decoder, so that model can utilize accurate position information, guide decoder to extract more pure, more comprehensive micro target content features, effectively improve the flexibility and accuracy of target detection.Finally, the class information and position information are output by prediction head.The present application fully releases the potential of cloud computing power while solving the problem of missed detection in dynamic scene, suitable for high-precision intelligent supervision task in complex scene, provides effective technical support for safety hazard management.

The application discloses a small sample face recognition method based on prototype calibration and adaptive fine-tuning, comprising the following steps: 1) pre-training the encoder on a large-scale face database, capturing rich semantic information and class distribution, constructing an evaluation set, and ensuring that the training identity information does not overlap with the constructed evaluation set; 2) initializing the classifier weight through the weight imprint method, ensuring smooth transition, and inducing the discriminative structure of the encoder; 3) calculating the cosine similarity between the fine-tuning class prototype and the pre-training prototype, fusing the pre-training class information through weighted average, and calibrating the biased fine-tuning class prototype; 4) introducing an adaptive margin loss function, adjusting the sample penalty strength, and selectively fine-tuning the BatchNorm layer and the classifier weight; and 5) using the field-aware similarity NAC to calibrate the unknown class rejection, enhancing the similarity relationship with the field sample, and obtaining the final matching score. The method can improve the recognition accuracy and rejection ability of the model in the small sample learning scene.

The application discloses a specific behavior recognition method in a dense crowd environment, comprising the following steps: S1, data set acquisition, wherein the data set comprises a pedestrian detection data set and a behavior recognition data set; S2, data set preprocessing; S3, inputting images in the preprocessed pedestrian detection data set into a feature map pyramid network to extract corresponding features, generating a candidate region and class information of the candidate region through a region generation network; S4, removing overlapping targets by using a maximum value suppression algorithm; S5, performing specific behavior target recognition based on a classification recognition network of a residual network; S6, training grid parameters of the classification recognition network; S7, obtaining optimal grid parameters through step S6, and importing the optimal grid parameters into the classification recognition network, and testing through a behavior recognition data set. The method realizes the detection and recognition of specific behaviors in a dense crowd environment by fusing two stages of detection and recognition tasks for the special environment of the dense crowd.

The invention provides a method, a device, equipment and a medium for generating a code call chain based on a graph database, and the method comprises the steps: carrying out the static analysis of a target source code, so as to extract related information, which comprises class information, method information and a call relation; target nodes and target relationships are constructed in the initial graph database according to the related information to obtain a target graph database, the target nodes comprise class nodes, method nodes and interface nodes, and the target relationships comprise a first relationship between the class nodes and the method nodes and a second relationship between the method nodes; a third relationship between the class nodes and a fourth relationship between the class nodes and the interface nodes; and constructing a visual code calling chain according to the target node and the target relationship. Therefore, the related information of the target source code can be accurately and intuitively indicated based on the graph database, and the visual call chain view is generated.

The invention belongs to a computer vision and semantic segmentation technology, and discloses a weak supervision semantic segmentation method based on adaptive missing class correction, which comprises the following steps: training a classification model by using an image-level label, generating a CAM (Computer Aided Manufacturing), and generating a pseudo-label training semantic segmentation network model by using the CAM; the prediction result of the semantic segmentation network is analyzed, and the category of significant missing in the prediction result is counted; performing threshold segmentation on the CAM of the missing class by using the activation intensity, and constructing a position information mask of the missing class; and finally, in combination with the missing class position mask and missing class information, constructing a correction loss function for the missing class so as to improve the discrimination capability of the model for the missing class. According to the method, on the premise of not depending on a pixel-level segmentation label, the spatial position information in the CAM is fully utilized to provide additional weak supervision guidance for the model, and the quality of the pseudo label and the performance of the segmentation model are effectively improved.

To provide an equipment management system, a communication adapter, an equipment management method and a program for reducing the influence of the restriction of the storage capacity of equipment information.SOLUTION: A communication adapter 1 includes an apparatus information notification unit 114 which generates apparatus information notification information including apparatus information associated with the same time information about an illumination device 3 and a sensor 4 belonging to the same class, transmits the apparatus information notification information to a cloud server 2, and deletes apparatus information stored in an apparatus information buffer 121 each time the apparatus information notification information is transmitted. The cloud server 2 includes the equipment information acquisition unit 212 that, upon acquiring the equipment information notification information, stores the equipment information and the time information included in the equipment information notification information in association with each other in the equipment information buffer 221, and the equipment information aggregation unit 213 that, based on the class information, aggregates the equipment information using the equipment information associated with the same time information stored in the equipment information buffer 221.SELECTED DRAWING: Figure 3

This invention discloses a multimodal balanced learning method based on label reshaping. Starting from label-side design, this invention proposes for the first time to balance multimodal learning by reshaping the cross-modal label space. A reshaping matrix is ​​generated using the prediction output of complementary modalities to inject cross-modal inter-class information. Simultaneously, the reshaping intensity is adaptively calculated based on the confidence differences between modalities to control the degree of label space transformation, thereby balancing the mapping difficulty from the feature space to the label space for different modalities. Finally, a target parameter update strategy is employed, applying different optimization objectives to different parts of the model to ensure training stability. This method dynamically reshapes the label space corresponding to each sample and each modality, constructing a balanced soft supervision signal rich in inter-class relationships. This balances the learning difficulty of different modalities at the source and injects cross-modal inter-class information to enhance the model's discriminative ability.

The application provides a memory fault processing method and device, a storage medium and an electronic device, the method comprising: in response to a memory fault processing instruction, obtaining a heap memory dump file generated by a virtual machine corresponding to the memory fault processing instruction when a memory fault occurs; determining class information of each java class in the heap memory dump file; determining a target java class in each java class according to the class information of each java class; determining a target instance object occupying the most memory and an associated instance object referenced by the target instance object in all instance objects of the target java class; and analyzing the target instance object and the associated instance object to determine fault information having an association relationship with the memory fault. The method provided by the application can quickly and accurately determine factors causing the memory fault.

The present application relates to a kind of multi-class weld defect classification method, system, equipment and medium based on the interaction of Transformer and channel, comprising: input image pre-processing, respectively to the input support image and query image division, embedding, position coding and class information embedding;Image block embedding sequence containing position and class information is transported Transformer encoder, and the embedding features of the image block of each channel containing class-related information and original are extracted;Class-related embedding vector and global visual context vector are used to adaptively adjust the embedding features of the image block of each channel;By calculating the similarity of support image and query image, judge whether they belong to the same class or defect type;The defect class of image is output through classification network;The present application effectively improves the precision of weld defect classification, especially in the scene of diverse defect types and small sample size, overcome the class confusion and small sample learning problem faced by traditional method.

A method for training an image classification model includes obtaining first prediction class annotation information of a first image by using an image classification network based on a first model parameter of an offset network being fixed; determining a second model parameter corresponding to the image classification network by using a classification loss function based on the image content class information and the first prediction class annotation information; obtaining second prediction class annotation information of the first image by using the offset network based on the second model parameter of the image classification network being fixed; determining a third model parameter corresponding to the offset network by using the classification loss function based on the image content class information and the second prediction class annotation information; and training a semantic image segmentation network model based on the second model parameter and the third model parameter.

The invention discloses an ML-ELM-AE target motion pattern recognition method containing class information, and belongs to the technical field of situation cognition. An ELM-AE model is constructed, and an ML-ELM-AE model is constructed by stacking the ELM-AE model; carrying out the mapping of the motion features of the target through the ML-ELM-AE; a CELM classification model is constructed, and according to a difference vector set formed by inter-class samples, the weight of an input layer of the CELM classification model to a hidden layer and a hidden layer node bias item are optimized, so that the mapping of the samples from a feature space to a class space has regularity, and the precision and generalization ability of target motion pattern classification are improved.

A learning device according to the present invention comprises: an input feature expansion unit that generates, from input features which are inputs from a plurality of data sources, a plurality of similar input features which are similar to the input features; an input feature embedding unit that embeds each of the plurality of similar input features in a low-dimensional vector representation to generate embedded vectors; a dataset feature extraction unit that, for each of the plurality of data sources, extracts a dataset feature by aggregating vector representations of each class calculated from the input features; a class information inference unit that, on the basis of the dataset features and the embedded vectors, infers class information expressing labels of the input features; a mock input feature generation unit that uses a generative model to generate mock input features from the class information; an input feature identification unit that predicts, as identification results, class information having the closest vector representation to the embedded vectors among the class information; and a model training unit that generates a trained model using the embedded vectors, the mock input features, and the identification results as inputs.

The invention relates to the technical field of data processing, in particular to a health management information prompting method and device based on wearable equipment, and the method comprises the steps: collecting first-class information and second-class information of a target object through the wearable equipment, and obtaining basic health information of the target object; the first type of information is used for representing physiological signals of the target object; the second type of information is used for representing the pulse waveform and the body surface state of the target object; the basic health information is used for representing historical body health examination information of the target object; analyzing the first type of information and the basic health information to obtain a first analysis result; analyzing the second type of information and the basic health information to obtain a second analysis result; based on a pre-trained multi-modal feature fusion model, performing fusion analysis on the first analysis result and the second analysis result to generate a comprehensive analysis result representing the physiological state of the target object; and according to the comprehensive analysis result, the wearable device carries out health management information prompting on the target object.