69results about How to "Reduce the need for computing resources" patented technology

The invention relates to a character-driven lip sound synchronous digital human generation method and device, equipment and a medium, and the method comprises the steps: obtaining an input text and a single object video, and carrying out the preprocessing of the input text, and obtaining a phoneme sequence and semantic features; the method comprises the following steps: preprocessing a single-object video to obtain a circularly played video paragraph, a mask face image and a reference face image; generating voice features based on the phoneme sequence and the semantic features; generating an audio sequence based on the voice features; acquiring face deformation parameters according to the mask face image, the reference face image and the voice features, and deforming the reference face image to generate a target face image; and embedding the target face image into a video frame of the video paragraph to generate a video sequence, and splicing the video sequence with the audio sequence to obtain a target digital human video. The digital human video is generated based on text driving, so that the voice and lip-shaped video matched with the text can be generated in time, the time delay for waiting for text to be converted into voice processing is eliminated, and the digital human video production efficiency is greatly improved.

The invention discloses a classroom teaching quality auxiliary evaluation method, device and system based on cloud edge collaboration, and the method comprises the steps: obtaining video image data, collected by a camera, of students in a classroom in real time, and transmitting the video image data to an edge device; and in the edge device, performing student classroom behavior identification and tracking on the received video image data by using a pre-trained student classroom behavior identification model, and fusing the identification results of adjacent multiple frames with the identification result of the current frame in the tracking process to obtain the final identification result of the current frame. Evaluating the classroom teaching quality according to the student classroom behavior recognition result, wherein the student classroom behavior recognition model is obtained by training based on a multi-label target detection algorithm and by adopting a knowledge distillation structure in advance; and sending a student classroom behavior identification result and an evaluation result to a cloud video server for storage. The method has the advantages of being simple in implementation method, low in cost, high in real-time performance and precision, high in robustness and the like.

The invention discloses a customer voice analysis system based on a large model, and the system comprises a voice data collection and preprocessing module which is used for collecting and preprocessing customer voice data; the voice recognition module is used for recognizing the preprocessed customer voice data; the feature fusion module is used for extracting semantic features of the text data and fusing customer voice data features to generate voice text representation; the large model fine tuning module is used for performing fine tuning on the pre-trained starry fire cognition large model; the semantic analysis and emotion recognition module is used for processing the voice text representation, extracting semantic information, recognizing an emotion state and generating an analysis result; and the dialogue context management module is used for associating customer historical voice data, establishing dialogue context information and updating a voice analysis result. According to the method, the stellate fire cognitive large model and the improved quantitative low-rank adaptation are utilized, and the method has the advantages of high precision, low calculation cost and high intelligent adaptability.

The invention relates to the technical field of medical image processing, and particularly discloses a pathological tissue image-based tumor analysis method, which comprises the steps of data acquisition and preprocessing, network construction, image enhancement, adjacent layer feature fusion, image classification and knowledge distillation. According to the scheme, features are extracted from different levels by adopting an image enhancement and comparative learning method, the capability of extracting global features and capturing local details of the model is improved, and by calculating strong and weak enhancement difference loss, it is ensured that enhancement operation does not excessively change feature representation of the image, and overfitting of the model is prevented; features are extracted and fused from each layer of feature map of the teacher network through an attention mechanism, and the detail capturing ability of the model to a pathological image is enhanced; a knowledge distillation method is used, knowledge of a teacher network is introduced, and a student network not only pays attention to classification of hard tags, but also can capture a relative relationship between classes, so that the classification accuracy is improved, and the calculation resource demand of a model is reduced.

The invention discloses a TCBL wind field high-resolution LES simulation method facing gray area parameterization, a medium and a program product. According to the method, a TCBL idealized initial gradient wind field is constructed based on a modified Rankine vortex model, controllability and stability of initial conditions are ensured by specifying key kinetic parameters, and thermodynamic profiles under different wind speed conditions are obtained based on synthetic analysis of tropical cyclone sounding data. And correcting the bottom of the boundary layer by adopting a quadratic and logarithmic function fitting method. Through high-resolution grid configuration and dynamic boundary conditions, secondary grid turbulence is processed in combination with an SGS-TKE closed model, and adaptability of the model to strong rotation and high wind shear environments is optimized by adopting a dynamic correction mechanism. Besides, the stability of thermodynamic variables is maintained through large-scale tendency correction, and high-resolution three-dimensional wind field and turbulence statistical reference data are output and used for verifying and improving a turbulence parameterization scheme of the TCBL under gray region resolution.

The invention discloses a control method of a mechanical arm, the mechanical arm and a computer readable storage medium. The method comprises the steps of receiving a task instruction and an environment image; wherein the environment image has depth information; inputting the task instruction and the environment image into a large language model, so that the large language model outputs natural language instructions corresponding to the sub-tasks in sequence; wherein the large language model does not have code generation capability; wherein each subtask is predicted when the execution of the previous subtask is completed; and the natural language instructions are sequentially input into the mechanical arm control model according to the sequence, so that the mechanical arm control model controls the mechanical arm to execute corresponding actions based on the natural language instructions until all the subtasks are completed. By means of the mode, the problem that in the related technology, understanding of action semantics is insufficient can be solved.

The invention discloses an improved YOLOv8 fish image recognition method based on transfer learning, and the method comprises the following steps: collecting a fish image data set, and carrying out the preprocessing of the data set; correspondingly marking category names on the fish image data set according to fish categories, and dividing the fish image data set into a training set, a verification set and a test set in proportion; a standard YOLOv8 model is improved; all fish images in the ImageNet data set are adopted to pre-train the improved YOLOv8 model, and obtained weight parameters are migrated to the model; performing optimization parameter training on the improved YOLOv8 model after transfer learning by using the training set and the verification set; and performing target detection on the test set through the optimized improved YOLOv8 model, and outputting a fish identification result. According to the method, the feature extraction and recognition precision is effectively improved, the computing resource demand is reduced, and the robustness and adaptability of the model are enhanced; according to the method, the learning efficiency and the recognition precision of the model are remarkably improved, data dependence is reduced, and the model has high generalization ability.

The invention discloses a current inversion method based on a magnetic sensor array, and the method comprises the steps: S1, setting a plurality of measurement points, uniformly dividing a conductor layout region in a to-be-measured part of a to-be-measured multi-conductor system into grids with M * N nodes, and numbering each grid node as (i, j); constructing a coordinate system and recording coordinates of all grid nodes; s2, establishing a finite element simulation calculation model, calculating a reference position coefficient matrix of all measurement points, and constructing a reference position coefficient matrix database; s3, calculating a conductor position coefficient matrix, and constructing a position coefficient matrix; s4, according to the magnetic field measurement value acquired by the magnetic sensor array, the conductor current is calculated through inversion; and S5, after resampling the magnetic field measurement value, substituting the updated measurement value and the position coefficient matrix stored in the step S4 into the linear equation set constructed in the step S4 to solve the current of each conductor. The method has the characteristics of rapid calculation and low operand while the accuracy is ensured, and the inversion calculation efficiency can be improved.

The invention provides a multi-classifier fusion method based on a quantum convolutional neural network, and aims to solve the problem of large deep learning computing resource demand and the problem of poor multi-classifier fusion effect, the method is based on a parameterized quantum circuit, the entanglement capability and expressible capability of the circuit are considered, and the multi-classifier fusion effect is improved. Three kinds of quantum convolutional neural networks with different convolutional structures are provided as base classifiers, required quantum bits and training parameters are reduced, and meanwhile, high classification accuracy and certain difference are achieved; according to the method, the output result of each base classifier is converted into an evidence form, and an improved quantum average combination method is used as a fusion method of multiple quantum classifiers, so that conflicts among the base classifiers are solved, and accurate classification of a multi-classifier fusion system is realized.

The invention relates to the field of network live broadcast, and discloses a live broadcast stream image segmentation method and device, equipment and a medium, and the method comprises the steps: sequentially calling image frames from an image sequence of a live broadcast stream as current image frames; a depth estimation segmentation model is called, a segmented figure mask corresponding to the figure image is determined based on the current image frame, and the depth estimation segmentation model is supervised and trained by using scene depth information; performing inter-frame smoothing processing on the segmented figure mask of the current image frame by using the practical figure mask iteration of the historical image frame to obtain a practical figure mask of the current image frame; and determining a figure image in the current image frame according to the practical figure mask of the current image frame. By fusing the depth information and the inter-frame smoothing technology, real-time, accurate and stable live broadcast stream image segmentation is realized, the live broadcast watching experience is improved, and the industry technology progress is promoted.

The invention discloses a road surface disaster detection method and system, electronic equipment and a storage medium, and relates to the technical field of intelligent traffic management and computer vision, and the method comprises the steps: collecting image data containing a plurality of road surface disaster types, carrying out the data marking in each image data, and carrying out the data marking; a data sample is constructed through each piece of image data subjected to data annotation; training the improved YOLOv10 model by using the data sample until a loss function of the improved model reaches a preset end condition, and determining the improved model of which the loss function reaches the preset end condition as a disaster detection model; acquiring a to-be-detected image of a to-be-detected road surface, and inputting the to-be-detected image into the disaster detection model for processing to obtain a disaster detection result of the to-be-detected road surface, the disaster detection result including a disaster type and a disaster position; by combining the deep learning technology and optimization algorithm design, the problem that the existing method is insufficient in detection precision, robustness and applicability is solved.

The invention discloses a low-resolution face recognition method and device, electronic equipment and a medium, and the method comprises the steps: obtaining a plurality of face image sample pairs which comprise high-resolution face images and corresponding low-resolution face images; constructing a teacher model and a corresponding student model, and training the teacher model according to the high-resolution face image to obtain a face recognition teacher model; respectively inputting the face image sample pair into a face recognition teacher model and a student model, and enabling the student model to learn the characterization capability of the face recognition teacher model through knowledge distillation to obtain a face recognition student model; and deploying the face recognition student model to an end side device, and performing face recognition through the end side device. According to the method, the face recognition precision of the low-resolution image is improved, the bandwidth consumption and the server burden are reduced, the real-time performance and the reliability of face recognition are improved, and the method can be widely applied to the technical field of artificial intelligence.

The invention relates to the technical field of industrial automation and intelligent detection, and discloses a blast furnace crude fuel granularity intelligent detection system and method based on a deep learning algorithm. Firstly, an image of blast furnace crude fuel is collected in real time through the image acquisition module, and then the image is subjected to denoising and enhancement processing through the preprocessing module. Then, an improved YOLOv8 model which takes the OfficientNet as a backbone network and combines an FPN and PAN feature fusion strategy is adopted to carry out granularity detection, bounding box regression is optimized through a GIoU Loss loss function, and the detection precision and speed are remarkably improved. In addition, the system also has a particle size distribution statistical function, and can generate a particle size distribution histogram in real time. Therefore, intelligent upgrading of blast furnace crude fuel granularity detection is realized, the production efficiency and the product quality are improved, and the requirements of modern steel production on efficient, accurate and safe granularity detection are met.

The invention discloses a built-in lightweight new energy adaptive detection method, and relates to the technical field of object detection, and the method comprises the steps: S1, constructing a database; s2, preprocessing the data; s3, constructing a model architecture; s4, integrating attention modules; s5, preparing self-supervised comparative learning; s6, training early-stage preparation; step S7, performing self-supervised contrast learning training; step S8, performing unsupervised adaptive training; s9, optimizing the model and adjusting parameters; and S10, verifying and evaluating the model. By adopting the built-in lightweight new energy self-adaptive detection method, the real-time monitoring capability and detection precision of the system are improved, the robustness of the system under variable weather and environment conditions is enhanced, the operation efficiency and experience of a user are improved, and the user experience is improved. And a new technical path and a new solution are provided for sustainable development of small hydropower, photovoltaic and wind power projects.

The invention discloses a personnel abnormal behavior analysis method based on deep learning, and belongs to the technical field of image data processing, and the method comprises the steps: obtaining a personnel abnormal behavior detection data set D; constructing an improved YOLOv5 network based on the spatial feature enhancement module; training the improved YOLOv5 network to obtain a target detection model; based on the loss functions of the improved YOLOv5 network, the DeepSORT algorithm and the improved SlowFast network, training the SlowFast network to obtain a SlowFast model; and carrying out abnormal behavior identification on a target in the to-be-identified video sample based on the target detection model, a DeepSORT algorithm and a SlowFast model. According to the method, the problem of class imbalance is solved by using adaptive weight, and the accuracy of video action recognition is improved by using time information, so that the robustness and accuracy of a model to a complex scene are improved, the computing resource demand is reduced through lightweight design, and greater flexibility and efficiency are provided for practical application.

The invention provides a multi-step building cold load prediction method based on deep time-space feature fusion learning, and the method comprises the following steps: S1, obtaining time-space historical data related to building cold load prediction, including outdoor climate condition data, building cold load data, cold source system topological structure information and control state information; s2, data cleaning and data preprocessing are carried out on the existing historical time sequence operation data, the data cleaning comprises anomaly detection processing, and the data preprocessing comprises data smoothing and normalization processing; s3, establishing a multi-step building cold load prediction model CRG-Informer of depth time-space feature fusion learning based on existing data, wherein the model comprises a graph model establishment and embedding process and an Informer model prediction process; and S4, building cold load prediction is carried out by using the trained CRG-Informer model and real-time building operation and maintenance data. According to the method, the accuracy of building cold load prediction can be improved, and the building energy-saving effect is improved.

The invention discloses a pre-training visual language model-based cross-country scene unmanned vehicle auxiliary decision-making method, and belongs to the technical field of unmanned driving. Comprising the following steps: constructing a simulation data set for training a three-dimensional target detection network; carrying out cross-modal consistency fusion and cue word design; and generating and mapping driver element actions and decisions. According to the invention, by fusing a three-dimensional target detection network and a pre-trained visual language model, introducing a thinking chain type reasoning and selective spatial information injection mechanism, and combining a trigger type reasoning process and meta-action sequence generation, high-level understanding and auxiliary decision making of a cross-country scene are realized; therefore, the defects of a traditional geometric programming algorithm in an unstructured environment are overcome.

The invention provides an abnormal cut tobacco identification and detection method and device based on a YOLO model, and belongs to the technical field of cigarette manufacturing. The abnormal tobacco shred identification and detection method comprises the following steps: acquiring an abnormal tobacco shred image and constructing a data set of the abnormal tobacco shred image, wherein the abnormal tobacco shred image takes densely stacked normal tobacco shred as a background; an abnormal tobacco shred recognition model based on a YOLO model is constructed, the abnormal tobacco shred recognition model at least comprises a backbone network, a neck network and a head network, and a BiFormer attention mechanism based on a Transform architecture is introduced into the neck network; training the abnormal tobacco shred identification model through the data set; and identifying abnormal tobacco shreds in the tobacco shred pile by using the trained abnormal tobacco shred identification model. Through a deep learning algorithm, the identification precision is remarkably improved, and abnormal tobacco shreds with the density similar to that of normal tobacco shreds are effectively distinguished.

The invention discloses a risk identification method and system in a cement fiberboard construction process, and the method comprises the steps: extracting a construction personnel wearing reference image which is marked as a wearing standard from a database, and taking the image as a detection reference; and carrying out multi-scale feature extraction and fine-grained semantic matching analysis on the construction personnel wearing reference image and the construction personnel wearing image acquired on site by adopting an image processing technology based on deep learning so as to intelligently judge whether the construction personnel wear normatively or not and carry out corresponding early warning prompt. In this way, the computing resource demand can be effectively reduced, and the efficiency and accuracy of construction site safety management are improved.

This invention relates to a physical information neural network-based intelligent performance prediction method and system for automated multi-material fiber placement on complex curved surfaces in aircraft. The invention includes: collecting fiber placement process parameters; performing standardized preprocessing and physical constraint-enhanced sampling to obtain an enhanced training dataset; constructing a high-order nonlinear sparse regression candidate library based on the enhanced training dataset; jointly identifying piecewise smooth physical differential equation systems and their mode switching logic using sparse Bayesian regression and Hidden Markov Models to obtain an embeddable inverse mechanism model; constructing a Physical Information Embedded Generative Adversarial Network (PI-GAN); collaboratively optimizing the generator and discriminator through adversarial training to output surface defect indices and mechanical performance indices; and initiating incremental self-learning when introducing new process scenarios to achieve model self-evolution. This invention achieves end-to-end accurate prediction of multi-dimensional quality indices such as surface defects and mechanical performance.

The invention discloses an index tensor vectorization calculation method and a related product. The method comprises the following steps: acquiring the length of a first index tensor of a first sparse tensor and the length of a second index tensor of a second sparse tensor; based on the length of the first index tensor, obtaining at least one first target instruction corresponding to the first index tensor from an instruction set; obtaining at least one second target instruction corresponding to the second index tensor from the instruction set based on the length of the second index tensor; and executing the at least one first target instruction and the at least one second target instruction to perform vectorization calculation on the first index tensor and the second index tensor to obtain a third index tensor. According to the invention, the computing resources of the computing units can be saved and the splicing efficiency of the sparse tensor can be improved.

The invention discloses a product recommendation method, system and device and a storage medium. The method comprises the steps of obtaining a target data set; performing two-way feature extraction on the target data set to obtain user features and financial product features after feature coding; the user features and the financial product features are respectively processed based on the multi-head self-attention layer, and the internal correlation of the user features and the financial product features is enhanced; fusing the user features and the financial product features by adopting a multi-head cross attention layer to obtain a first interaction feature sequence after the user feature data influences the financial product feature data and a second interaction feature sequence after the financial product feature data influences the user feature data; and based on the first interaction feature sequence, the second interaction feature sequence and the statistical pooling layer, determining a recommendation result of the model. The fused feature vector comprehensively contains the advantages of the user features and the financial product features, and the accuracy of the financial product recommendation model is improved.

The invention provides a method for rapidly detecting microbial species, and particularly, the method comprises the following steps: constructing a double-level database and a corresponding gene cluster; constructing a reference matrix X based on the database and the gene cluster; comparing to-be-detected third-generation sequencing data with the gene cluster to generate equivalence classes, and constructing a counting matrix Y based on the equivalence classes and the gene cluster; based on the counting matrix Y and the reference matrix X, determining the gene abundance in the to-be-detected sample; and determining microbial species in the sample to be detected based on the gene abundance. The rapid pathogen detection method provided by the invention is high in analysis speed, fully utilizes the advantages of the three-generation sequencing data, and can more directly obtain the microorganism composition information of the sample; the method is low in computing resource demand, easy to deploy and high in precision detection capability.

The invention provides a pavement icing visual detection method based on deep learning, and the method comprises the steps: S1, obtaining pavement images in different states, and forming a pavement image data set; step S2, constructing and training an improved Resnet-18 model; constructing a feature extraction module, wherein the feature extraction module comprises an average pooling layer, a convolution layer, four parallel cavity convolution layers, a first feature fusion module, an average pooling layer and a convolution layer which are connected in sequence; a feature extraction module is used for enhancing shallow features extracted by the pre-trained Resnet-18 model and then fusing the shallow features into deep features extracted by the pre-trained Resnet-18 model, so that an improved Resnet-18 model is obtained; and S3, processing the road surface image data by using an improved Resnet-18 model to obtain probabilities of different road surface categories. The system has the advantages of high detection precision and high intelligent degree.

The application provides a method and system for predicting main steam parameters before steam warming of a combined cycle unit, the method comprising the following steps: obtaining historical operation data of a target power generating unit, the historical operation data comprising a plurality of observable input parameters and output parameters closely related to a main steam system; constructing an NARX neural network model, training the neural network model based on the historical operation data, and obtaining a trained main steam parameter prediction model; deploying the main steam parameter prediction model on a general computing device; obtaining initial state parameters under a current working condition before starting the power generating unit, and inputting the initial state parameters into the main steam parameter prediction model; and outputting a predicted trajectory of main steam key parameters before the high-pressure bypass system is put into operation from the main steam parameter prediction model.

This invention discloses a method and system for identifying innovative points in academic papers, relating to the field of artificial intelligence technology. The method first processes the paper abstract through a concept path extraction module, including: structured semantic segmentation, concept pair extraction and verification, constraint relation triple generation, hierarchical verification, and path optimization, to construct a complete set of concept paths for the paper. Then, in a rare path discovery module, the popularity index of each concept path in the global path frequency dictionary is calculated and compared with a threshold to identify rare paths as the scientific innovative points of the paper. This invention effectively suppresses the illusion of language models through a knowledge graph constraint mechanism, significantly improving the accuracy of concept path extraction and its coverage of long-tail concepts, achieving accurate and efficient identification of innovative points in academic papers at the path level.

The application discloses a scene generation method and device based on source-load uncertainty and equipment, and relates to the technical field of energy scheduling. The method comprises the following steps: acquiring real-time source-load data of a power system; processing the real-time source-load data based on a target prediction model to obtain a multi-time-scale source-load prediction result, wherein the target prediction model is obtained by compressing the weights of a long short-term memory network prediction model; generating a collaborative scheduling strategy of source-load resources under the multi-time-scale source-load prediction result based on a multi-agent scheduling decision model; and generating a hydropower source-load scene set of the power system at different time stages under the source-load uncertainty based on the collaborative scheduling strategy. In the foregoing manner, the prediction link and the decision optimization link are closely coupled, the multi-time-scale prediction result is used to dynamically guide the reinforcement learning decision, the adaptability of the system to the source-load uncertainty is enhanced, and the organic collaboration of the multi-time-scale and the accurate generation of the uncertainty scene are realized.

The application discloses a resource self-adaptive federated learning multi-modal knee osteoarthritis auxiliary diagnosis method, which comprises the following steps: a server initializes a multi-modal model for knee osteoarthritis diagnosis, and pre-deploys a low-rank adaptive plug-in in a trainable layer; after a client reports local hardware resources, the server allocates a resource echelon and issues a two-dimensional clipping ratio; the client clips the model in depth and width based on the clipping ratio, activates a LoRA module of a specified layer, loads local knee osteoarthritis X-ray images and clinical text data for multi-modal training, and establishes semantic mapping of image features and clinical descriptions; the client only uploads an activated LoRA parameter update amount, the server aggregates and optimizes a global model through weighted average aggregation and key layer compensation; and the learning ability of the model to key discriminative features of knee osteoarthritis is strengthened in combination with real-time resource monitoring and knowledge distillation. The application covers heterogeneous computing power devices through a resource self-adaptive mechanism, and improves the participation efficiency of primary medical institutions.

The application discloses a kind of multi-energy distribution network voltage control methods, comprising the following steps, define local voltage control rule;Analysis distributed power output causes voltage fluctuation upper and lower limit;Build global voltage control function;Algorithm is solved to objective function, the local voltage control rule is acquisition control point voltage u, determine whether to satisfy the following constraint, set the reactive power output of voltage compensation device.The multi-energy distribution network voltage control method, considering the uncertainty fluctuation of distributed power supply, improve the accuracy of voltage fluctuation analysis, formulate the localized voltage adjustment rule and global voltage optimization function, realize passive control and active control combination, reduce fluctuation rate, using interval number to express the uncertainty of distributed power supply, less computing resource demand, control reaction speed is fast.

The application discloses a multi-agent material calculation consistency calibration method, belonging to the cross field of material science, artificial intelligence and computational physics. The method adopts a master-slave structure multi-agent system, containing one master agent and at least three slave agents. The master agent is responsible for overall coordination, embedding physical mechanism constraints, consistency calibration and output results. The method steps are: constructing a calculation system, introducing physical mechanism constraints and embedding slave agent calculation models, collecting calculation data, performing bias analysis, adjusting parameters or correcting results through iterative calibration, and finally verifying and outputting the final results. The application introduces physical mechanism constraints, combines bias analysis and iterative calibration, and solves the problems of poor consistency of existing multi-agent material calculation results and easy violation of physical laws. Experiments show that the method reduces the deviation of the calculation results and the experimental measured values to below 3%, improves the calibration efficiency by more than 40%, and improves the calculation accuracy by more than 35% compared with the prior art, thereby improving the accuracy and engineering practicability of high-throughput calculation of material performance.