28results about How to "Improve inference accuracy" patented technology

The invention belongs to the technical field of artificial intelligence application, and particularly discloses a full-stack autonomous controllable artificial intelligence reasoning method and terminal based on a Feiteng processor, and the method comprises the following steps: A, initializing hardware; b, pruning and compressing the model; c, converting and optimizing the model; d, compiling and deploying the model; e, data preprocessing; f, performing reasoning execution; and step G, post-processing the reasoning result. The method has the beneficial effects that optimization is carried out aiming at intelligent management requirements of new energy power generation and traditional power generation, the YOLOv5 deep learning algorithm is integrated, and optimization processing such as quantification, pruning and weight compression is carried out on the YOLOv5 deep learning algorithm, so that the reasoning precision is remarkably improved, and the system delay and power consumption are effectively reduced. The system has the functions of real-time inspection and safety control, can efficiently monitor the power generation equipment, remarkably improves the intelligent level of operation, reduces unpredictable factors of manual operation, and improves the production efficiency and safety.

The invention discloses an article label identification method based on fuzzy logic control. The method comprises the following steps: S1, data acquisition; s2, preprocessing the image data and the environment data; s3, performing classification according to an improved clustering algorithm to generate an initial fuzzy rule base; s4, using an improved genetic algorithm to optimize the initial fuzzy rule base, adopting adaptive variation and crossover operation, introducing an elitist retention mechanism, designing a multi-objective optimized fitness function, and generating an optimized fuzzy rule base; s5, constructing a three-layer fuzzy system; s6, fusing three layers of reasoning results through a fuzzy operator to obtain a final label identification result, and if the reasoning confidence is relatively low, performing reasoning again; and S7, storing related data, and if the user corrects the identification result, automatically triggering the improved genetic algorithm to update the fuzzy rule base and the reasoning parameters. Article label identification is optimized through multi-level fuzzy reasoning and the improved genetic algorithm, and the method has the advantages of high adaptability, high precision and the like.

The present invention relates to the field of intelligent control technology, specifically to a digital control method and system for the online operation of multiple punching machines, which obtains production line historical data and real-time collection of punching machine operation status data, combines product appearance images and quality inspection parameters, realizes product appearance defect detection through the YOLOv7 model, generates offset quality inspection parameters in combination with quality inspection standards; and generates process abnormality positioning results using the offset quality inspection parameters and appearance defect detection results. The process abnormality positioning results are fused with the punching machine operation status data to form multimodal comprehensive operation feature data for identifying abnormal processes. After the abnormal process is detected, the abnormal propagation path is analyzed through a dynamic working condition adaptive graph neural network model, the current working condition characteristics are extracted, and similarity matching is performed with historical data to extract the best historical reference parameters. The production process adjustment parameters are generated by dynamically correcting the historical reference parameters. The adaptive ability and flexible response ability of the production line to complex working conditions are improved.

The model generation method according to the present disclosure is an information processing method executed by a computer, and includes machine learning that implements an inference model using a plurality of training images. The inference model includes a compression module and an inference module configured to infer a solution to a task for a sub-region in an input image. The compression module is configured to generate compressed information by compressing information about an extended region that includes the sub-region and is wider than the sub-region. The inference module is configured to derive the solution to the task from information about the sub-region and the compressed information obtained by the compression module.

The present invention discloses a method and device for detecting anomalies in high-dimensional network data, the method comprising obtaining high-dimensional data to be detected and a normal data set; preprocessing the normal data set to obtain a preprocessed data set; using the preprocessed data set to train a preset high-dimensional network data anomaly detection model to obtain the trained high-dimensional network data anomaly detection model; using the trained high-dimensional network data anomaly detection model to process the preprocessed data set and the high-dimensional data to be detected, respectively, to obtain a normal characterization variable set, a normal latent variable set, a characterization variable to be detected, and a latent variable to be detected; processing the normal characterization variable set, the normal latent variable set, the characterization variable to be detected, and the latent variable to be detected to obtain a detection result. The method of the present invention can improve the reasoning accuracy, detect abnormal data more accurately, and improve the accuracy of high-dimensional network data anomaly detection.

The invention relates to the field of visual target tracking and artificial intelligence, and provides a training method and device and a tracking method and device of a visual target tracking model, and the training method comprises the steps: obtaining a plurality of pairs of training samples; inputting the training sample into a feature embedding module to obtain template image features and search image features; utilizing a feature extraction module to execute feature extraction operation to obtain a first extraction result and a second extraction result; fusing the first extraction result and the second extraction result by using a feature fusion module to obtain a fusion result; performing prediction based on the fusion result by using the prediction head to obtain a prediction tracking result; training loss is obtained based on the prediction tracking result, the real labeling information and the distillation labeling information, the model is trained, and a trained visual target tracking model is obtained. The problem that an existing single-flow tracker has certain calculation redundancy and limits the reasoning speed is solved, and the precision of the model is improved while the high reasoning speed is reserved.

The invention discloses a target detection method based on foreground and background splitting and learning path feature distillation, a storage medium and equipment, and the method comprises the steps: collecting different types of pictures, and carrying out the target frame and target type marking of each picture; respectively inputting the marked picture into a teacher model and a student model, carrying out feature distillation based on foreground and background splitting and a learning path, and guiding the training of the student model according to a feature distillation loss function between the teacher model and the student model until a total loss function is converged, thereby completing the training of the student model; and inputting the pictures collected in real time into the trained student model for reasoning to obtain the category and position information of the target. The knowledge of the teacher model is migrated to the student model, so that the target detection performance of the student model approaches to the teacher model through foreground and background splitting and learning path feature distillation on the premise of not increasing the parameter quantity of the student model, and the method has the characteristics of low resource consumption, high reasoning speed and suitability for mobile equipment.

The invention relates to the technical field of silicon carbide defect detection, and discloses a silicon carbide product defect detection method and system based on an intelligent sensor, and the method comprises the steps: obtaining initial time sequence data; executing spatial feature mapping according to the data, extracting a preliminary boundary of a purity non-uniform region, and obtaining a purity deviation feature vector of the preliminary boundary; performing multi-time-domain evolution analysis based on the vector to obtain an optimized defect transition trajectory sequence; carrying out time domain differential analysis on the sequence, and extracting a dynamic evolution path index; the purity is judged, and an evolution influence quantization matrix is generated; further fusing the associated information to obtain a defect evolution influence factor; and generating a defect detection report based on the factor to obtain a final detection result. According to the method, defect dynamic evolution tracking under the high-temperature working condition can be achieved.

The present application provides a road traffic management method and system based on digital twins, which combines a first traffic flow mean statistical sequence of a target section in a first monitoring period and a second traffic flow mean statistical sequence of a control section in a second monitoring period, and then extracts characterization information to obtain a first characterization vector of traffic flow evolution characteristics of the same road where the target section and the control section are located, and then combines the first characterization vector and the first maintenance rate statistical sequence and loads them into a first inference network. The first inference network can combine the traffic flow evolution characteristics of the road and the actual maintenance time of the first maintenance rate statistical sequence to obtain an accurate target traffic flow mean. Because the target traffic state is divided into a target traffic flow mean and a traffic flow maintenance time, and each of them is inferred separately, and the inference accuracy of the target traffic flow mean and the traffic flow maintenance time is high, the inference accuracy of the target traffic state can be increased, which helps to achieve accurate data visualization.

This application provides a monolithic 3D integrated chip, relating to the field of semiconductor technology, primarily addressing the problem that current resistive random access memory (RRAM)-based logic architectures remain relatively fixed and lack flexible reconfigurability. The monolithic 3D integrated chip includes a logic layer, a memory computing layer, and a reconfigurable layer stacked sequentially, with interconnections between these layers. The logic layer transmits computation signals and data to the memory computing layer and configuration data and signals to the reconfigurable layer. The memory computing layer, under the control of the computation signals, performs matrix multiplication or addition on the computation data and outputs the computation results. The reconfigurable layer, under the control of the configuration signals, performs reconfigurable processing on the computation results according to the configuration data. This application endows the chip with flexible programmability and reconfigurability.

The present invention discloses a customized deep neural network model compression method and system based on cloud-edge collaboration. The method is based on a high-precision network model that has been trained for a long time in the cloud. Based on the personalized needs of users, the network model is classified into categories of concern and categories of no concern. Then, the categories of concern, the user's data set and the model compression ratio are combined. The model is lightweighted through a neural network model compression method based on knowledge organization to meet the needs of scenarios with limited node resources on the user's edge side, improve the speed of model reasoning, and ensure a certain model accuracy. The present invention is based on a model training method based on user-focused classification needs and knowledge distillation technology, and is verified based on a public data set in the scenario of image classification.

The invention relates to the technical field of biomedical signal processing, and discloses a calcium ion imaging signal peak inference method and system based on deep learning, and the method comprises the following steps: segmenting an original noisy calcium ion signal time sequence through interval sampling to generate a recovery sequence, and training a one-dimensional U-Net neural network model; processing the historical noisy calcium ion signal time sequence by using a one-dimensional U-Net neural network model to obtain a historical de-noised signal, and training a one-dimensional convolutional neural network model by using the historical de-noised signal; and denoising the original noisy calcium ion signal time sequence to be processed, and inputting the denoised original noisy calcium ion signal time sequence into the one-dimensional convolutional neural network model to deduce a neuron absolute spike rate sequence. According to the method, noise can be suppressed through self-supervised denoising without external truth value data, the signal-to-noise ratio is improved by using a two-stage framework, a neuron discrete discharge event is accurately solved, and the problems that the prior art depends on a truth value database and the inference precision is low under the low signal-to-noise ratio are solved.

The present invention proposes a small sample knowledge reasoning method based on a decoupled learning strategy, and realizes comprehensive reasoning feature extraction, candidate entity establishment probability calculation and best candidate entity prediction based on a small sample knowledge reasoning model, wherein the comprehensive reasoning feature extraction extracts the comprehensive reasoning features of the head entity based on the embedding vector of the head entity and the relationship embedding vector, including the embedded features and the small sample features; the candidate entity establishment probability calculation inputs the embedded features and the small sample features of the head entity into the reasoning scorer, and calculates the probability of the candidate entity being established under the embedded features and the small sample features; the best candidate entity prediction combines the progressive learning factor and the probability of the candidate entity being established under the embedded features and the small sample features, calculates the comprehensive probability of the candidate entity reasoning being established, and determines the candidate entity with the highest comprehensive probability value as the final reasoning result. The present invention provides comprehensive reasoning information for the reasoning score function, and improves the small sample reasoning accuracy.

The invention provides a deep counterfeit content detection system and method based on a multi-modal large model, and the method comprises the steps: taking a video clip as input, employing a phased Video Swin Transform as a backbone network, and extracting multi-scale deep space-time visual features in video data; a Mel-frequency cepstrum coefficient fragment is used as input, a cascaded ERes2Net module is used for constructing an audio feature extraction network, a multi-stage fusion mechanism is introduced, extracted audio features are injected into a visual feature extraction branch stage by stage, and fused features are formed; according to the method, features are aggregated through parallel space-time pooling and a time attention mechanism, then internal association of audio-visual features is deeply mined in a local region, among all regions and in three dimensions of local-global through a core local-global interaction module, and finally, a deep forging detection result is output through a full connection layer. According to the scheme, the optimal effect is achieved on the largest deep counterfeiting detection data set in the industry.

The present disclosure relates to a scene flow estimation apparatus, a scene flow estimation method, and a non-transitory computer-readable medium, the scene flow estimation apparatus including: a hierarchical feature detector that detects hierarchical features of point cloud data related to an image frame; the corrector corrects a hierarchical feature of the detected point cloud data, the bidirectional feature detector detects a bidirectional scene flow feature of the corrected point cloud data, re-inputs a detection value, and repeatedly detects the bidirectional scene flow feature, and the estimator estimates a scene flow of a next image frame by using the bidirectional scene flow feature.

The present disclosure provides a data processing method, comprising: performing enhancement processing on original data collected by an event information collection device according to at least one data enhancement transformation, to obtain at least one piece of enhanced data, wherein each piece of the enhanced data corresponds to at least one data enhancement transformation; obtaining at least one piece of initial prediction result corresponding to at least one piece of to-be-predicted data according to the at least one piece of to-be-predicted data, wherein the at least one piece of to-be-predicted data comprises the at least one piece of enhanced data, or one piece of the original data and the at least one piece of enhanced data; and obtaining a target prediction result according to the at least one piece of initial prediction result. The present disclosure also provides a data processing device, an electronic device and a computer readable medium.

This application relates to the field of network information technology, and in particular to an automatic fault work order processing method and related equipment based on a multimodal model. The solution includes acquiring several fault reports of various modal types uploaded by a target object; inputting these fault reports into a pre-trained multimodal model; if the task type is a fault report data processing task, mapping the fault reports of different modal types to a unified semantic space using the multimodal model to obtain several token sequences of the same dimension; concatenating these token sequences and decoding them to generate fault report processing flow information; if the task type is a fault report data understanding task, independently encoding the fault reports of different modal types using the multimodal model to generate fault question and answer information, and issuing the fault report processing flow information and fault question and answer information. This application aims to assist installation and maintenance personnel in troubleshooting network faults, facilitating rapid fault location and resolution, and improving installation and maintenance efficiency.

This specification provides a method for fine-tuning a code model, based on a code library containing multiple source codes and corresponding patch codes. The method includes: obtaining a first representation vector corresponding to any first source code among the multiple source codes; the first representation vector is determined by performing a first encoding on the text of the first source code and a second encoding on the first abstract syntax tree corresponding to the first source code; based on the first representation vector, performing a similarity-based search in the code library to obtain a target context; the target context includes several other source codes whose similarity to the first source code meets a first threshold and their corresponding patch codes; generating prompt text based on the first source code and the target context, inputting it into the code model to obtain an inferred patch; and fine-tuning the parameters of the code model based on the inferred patch and the first patch code corresponding to the first source code.

This application relates to an intelligent simulation system and method. The intelligent simulation system includes at least: a simulation configuration specification generation system for generating a simulation configuration specification based on user-input simulation tasks; the simulation configuration specification describes simulation configuration information; a simulation task resource inference system for inferring the computational resources required to execute the simulation task based on the simulation configuration specification; a simulation orchestration system for determining a target solver from multiple solvers based on the simulation configuration specification, so that the target solver executes the simulation task based on the computational resources; and a simulation fault repair system for generating repair instructions and a diagnostic report when the target solver fails to simulate. Through multi-system collaborative coupling, the intelligent simulation system enables users to trigger the simulation workflow through natural language interaction, achieving end-to-end automation from mesh generation to solution analysis. This significantly reduces the user's familiarity with the underlying software operations and enables standardized delivery of simulation expert engineering capabilities.

The present invention relates to the field of image processing technology, solves the technical problem that the detection performance of the current supervised method is low due to the scarcity of abnormal samples, and particularly relates to an unsupervised industrial composite material CT image abnormality inspection method, the method comprising the following steps: obtaining CT images of industrial composite materials and preparing anomaly detection data sets; constructing an unsupervised anomaly detection network CTAnoNet model that converts the industrial composite material anomaly detection task into an image reconstruction and discrimination task; and constructing a joint loss function L for the unsupervised anomaly detection network CTAnoNet model. By converting the anomaly detection task into an image reconstruction and discrimination task, the present invention can effectively overcome the problem that the effective samples are scarce and the abnormal pattern of the CT image is difficult to effectively model, and realizes accurate detection of abnormal patterns under complex backgrounds.

The present invention provides a multi-modal information fusion reasoning method and device for a remote sensing interpretation integrated machine, which relates to the field of remote sensing image processing technology, and can solve the problem of insufficient accuracy and credibility of interpretation of traditional single-modal information interpretation methods. The method includes: obtaining two remote sensing images to be inferred, the two remote sensing images are in the same scene and belong to different modes; inputting each remote sensing image into a dual-branch encoder for processing, and obtaining the general branch features and the exclusive branch features of each remote sensing image; using a feature fusion module, the two general branch features are fused, and the two exclusive branch features are fused; the fused general branch features and the fused exclusive branch features are summed and input into the downstream task detection head to obtain the reasoning results of the two remote sensing images in the scene. The present invention not only realizes edge deployment on the remote sensing interpretation integrated machine, but also realizes the fusion of multi-modal observation information, thereby improving the reasoning accuracy of ground observation.

This invention discloses an AI-driven big data fusion-based anti-misoperation interlocking system for wind farm booster stations, relating to the field of anti-misoperation interlocking technology for wind farm booster stations. It includes a data acquisition module that collects real-time electrical parameters, operating status, and environmental parameters of the equipment, and uploads them after preprocessing; a big data fusion storage module that categorizes and stores multiple types of data according to a hierarchical strategy, optimizing access and retrieval; an AI intelligent analysis and decision-making module that integrates multi-source data to generate precise interlocking strategies; an anti-misoperation interlocking execution module that executes interlocking in a three-level mode and provides status feedback; a communication interaction module that achieves internal and external data interaction through a dual-mode architecture; a status monitoring module that monitors the operating status of each module and identifies faults; and an emergency handling module that initiates a response mechanism and restores operation after self-checking following fault resolution. This invention relies on AI and big data fusion technology, dynamically adapting interlocking thresholds to equipment and environmental conditions; supporting unattended operation and intelligent maintenance, providing reliable assurance for the safe operation of wind farm booster stations.

The present application provides a method, device and computer system for detecting anomalies in traffic water scenes. The target scene anomaly detection network is obtained based on multi-level transfer learning. The computing efficiency and accuracy of the target scene anomaly detection network obtained in this way are enhanced, and the accuracy of scene anomaly detection can be effectively improved. Transfer learning includes multiple levels. The water scene image template library includes an unlabeled water scene image template library without real category labels and a labeled water scene image template library with real category labels. Different levels load one of the water scene image template libraries for debugging, and use the execution results of the control neural network as constraint information to optimize the network parameters of the scene anomaly detection network to be trained, so as to realize the debugging of the first-level network transfer learning; this level can improve the universality of the network based on a large number of unlabeled water scene image templates, and the reasoning accuracy is higher.

A method of the present invention is a method of inferring a content ratio of a component in a sample containing a component selected from K types, including: heating a sample set, ionizing resultant gas components sequentially, and observing mass spectra continuously; acquiring two-dimensional mass spectra of the respective samples from the mass spectra, and merging two or more of these spectra to acquire a data matrix; performing non-negative matrix factorization on the data matrix; correcting an intensity distribution matrix through analysis on canonical correlation between a base spectrum matrix and the data matrix; acquiring a feature vector from a corrected intensity distribution matrix and expressing the sample in vector space; defining a K-1 dimensional simplex and determining an end member; and inferring a content ratio of the component in the sample on the basis of the end member and the feature vector. It is possible to infer the composition of a component in an unknown mixture even from a mass spectrum acquired under an ambient condition.

A frequency determination apparatus (30) acquires a past dataset in which given frequency-specific time-series data indicating frequency-specific communication status of a wireless base station (10) based on a signal-to-noise ratio (SNR) over a predetermined monitoring period in time series is associated with frequency-specific evaluation value data indicating an evaluation value evaluating the communication status during a subsequent period following the monitoring period by frequency. The frequency determination apparatus (30) generates, using the past dataset, an evaluation value inference machine learning model that takes an input of the frequency-specific time-series data and outputs the frequency-specific evaluation value data. The frequency determination apparatus (30) inputs the frequency-specific time-series data with a given preceding period as the monitoring period to the evaluation value inference machine learning model to acquire an output of the frequency-specific evaluation value data with a future period following the relevant preceding period as the subsequent period, and determines a frequency to be used for the wireless communication based on the acquired frequency-specific evaluation value data.

The invention provides an AI-based real-time monitoring and predicting method and system for icing strength, an icing environment parameter representation array is an implicit representation of related parameters for determining the icing strength of a monitored area in a monitored period, and an icing strength parameter representation array is an implicit representation of the icing strength of the monitored area in the monitored period. If the distribution coding array represents the position of the array element in the monitored icing correlation representation array set, the icing intensity of the prediction period is inferred by combining an integrated icing correlation representation array set formed by the icing intensity parameter representation array, the icing environment parameter representation array and the distribution coding array; and the reasoning precision of the icing strength can be improved.

The present disclosure is a control method of an image signal processor for an artificial neural network, the method can be configured to include: a step of acquiring an image; a step of determining at least one image characteristic data corresponding to the image; and a step of determining an image correction parameter (SFR preset) for improving the inference accuracy of an artificial neural network model based on the at least one image characteristic data and the inference accuracy attribute of the artificial neural network model.