102 results about "Adversarial network" patented technology

The application discloses a battery data expansion and energy management closed-loop method for complex traffic flow, and relates to the technical fields of new energy vehicle battery management and artificial intelligence control. The application is aimed at the problems of battery high dynamic operation data scarcity under complex traffic flow conditions, poor generalization of existing energy management strategies, and open-loop data generation being separated from real vehicle control, and first constructs associated data sets of traffic flow dynamic conditions and corresponding battery operation responses; then constructs a conditional generation adversarial network introducing battery physical constraints, directionally generates battery response data and constructs an expanded data set; based on the expanded data set, a deep reinforcement learning energy management strategy is trained offline, and finally the strategy is deployed to a vehicle-mounted control system, and online closed-loop fine-tuning is performed through real vehicle operation data. The application improves the robustness of the energy management strategy under complex conditions and effectively reduces the energy consumption of the whole vehicle.

This invention discloses a method and system for real-time monitoring and optimization of biomass gasifier operating conditions based on machine vision and multi-source data fusion. The method includes: synchronously acquiring gasifier images and sensor data; dynamically generating a thermal field distribution map from temperature data to construct a multimodal dataset; pre-training a cross-modal Transformer and using the thermal field distribution map as a physical prior to guide attention to focus on the high-temperature region, obtaining a cross-modal encoder; constructing a repair adversarial network with a fixed encoder; and using a discriminator to fuse image features, sensor features, and the thermal field distribution map for adversarial collaborative discrimination to achieve unsupervised anomaly detection; inputting the anomaly region into an improved U-Net, initializing attention weights with the thermal field distribution map, and outputting a segmentation mask and morphological features; fusing morphological features and sensor data to output operating condition categories and triggering early warnings; and generating optimization parameters based on the early warnings and triggering incremental learning. This invention achieves accurate monitoring and adaptive control of gasifier operating conditions.

The present application relates to the technical field of image processing, and more particularly to a generative adversarial network processing system for radiology CT image reconstruction, which comprises a preprocessing module, a double-path generator module, an attention discriminator module, a consistency constraint module and an adversarial training and optimization module. The present application realizes high-quality and high-fidelity reconstruction of CT images from low-quality projection data, and improves the clinical reliability of the reconstruction results.

The application relates to the technical field of big data processing, in particular to a fracture treatment scheme operation scheme simulation system based on big data analysis, which comprises a data preprocessing module, an adversarial network module and a model training and optimization module; a plurality of heterogeneous teacher models are used to pretrain a general large-scale fracture treatment data set, and fracture treatment decision knowledge learned by the plurality of heterogeneous teacher models is transferred to a student model for transfer learning. The application overcomes the contradiction between the small number of comminuted fracture cases (sample sparsity) of a single hospital and the high individualization of medical resources and capabilities (resource difference) by fusing the strategies of generative adversarial network (GAN) sample expansion and heterogeneous teacher model knowledge distillation transfer learning.

This application discloses an image enhancement method, electronic device, and storage medium. The method includes: acquiring a first image, wherein the first image is a low-light image to be enhanced; inputting the first image into a target generator for processing to obtain a second image, wherein the second image is a low-light enhanced image; wherein the target generator is trained based on a generator in a generative adversarial network, the generative adversarial network including the generator and a discriminator, the generator including a semantic segmentation network, the semantic segmentation network being used to assist in extracting low-light features at different levels in the first image.

The application provides an AI algorithm-based big data system operation monitoring method and device, and relates to the technical field of communication monitoring. The method comprises the following steps: firstly, collecting real-time position data and performance data of a fixed command center and a plurality of mobile terminals; then, controlling a phased array antenna to scan and receive a return signal based on the real-time position data, so as to generate electromagnetic environment information; next, inputting the information into a generative adversarial network to generate abnormal area information; then, inputting the abnormal area information and the performance data into a graph neural network based on an AI algorithm to construct a correlation graph and calculate the correlation degree between the abnormal area information and the performance data; finally, determining a target area and a correlation node affecting communication according to the correlation degree, and generating an operation monitoring result of the system. The application improves the positioning accuracy and prediction ability of the problem of regional communication performance degradation caused by specific electromagnetic interference in the emergency command big data system.

This invention relates to a cross-subject EEG emotion recognition method based on adversarial networks and soft negative sample contrastive learning, comprising the following steps: preprocessing an EEG signal dataset; constructing a pre-trained model, wherein the pre-trained model includes an encoder, a projector, and a discriminator of a generative adversarial network using a contrastive learning method; training the pre-trained model; and using the pre-trained encoder to achieve cross-subject EEG emotion recognition. This invention effectively improves the encoder's ability to extract cross-subject generalization features by combining the encoder with the projector and discriminator for adversarial training; and designs a contrastive loss function SDLoss that incorporates soft negative samples, enhancing the encoder's ability to capture shared neural responses under the same video stimuli, resulting in more stable recognition accuracy and stronger generalization performance in cross-subject emotion recognition tasks.

An apparatus comprises a memory communicatively coupled to a processor. The processor is configured to receive a reference file from a reference repository, determine an annotation for a reference file, generate a real sample based at least in part upon the reference portion and the annotation, receive a training file from the training repository, obtain multiple evaluation commands configured to incorporate one or more randomized changes onto the training file, and generate an adversarial network sample based at least in part upon the training file and the evaluation commands. Further, the processor is configured to execute the machine learning algorithm to determine whether the adversarial network sample at least partially matches the real sample and train the one or more machine learning models using the real sample and the adversarial network sample in response to determining that the adversarial network sample at least partially matches the real sample.

The application provides a virtual reality processing incomplete human body scanning data completion generation method for fashion design, relates to the technical field of fashion design, and comprises the following steps: obtaining human body 3D scanning data required by fashion design, pre-processing the scanning data to obtain sparse incomplete human body 3D point cloud data; based on a trained generative adversarial network, the sparse incomplete human body 3D point cloud data is completed to obtain complete human body 3D point cloud data; the application adopts a generative adversarial network GAN and combines a fashion design scene to perform special adaptation optimization, proposes a GAN-C completion algorithm, solves the problem that existing generative AI completion methods are not adapted to the fashion design scene and the completion morphological distortion, the algorithm optimizes a loss function L1 and network parameters by specially constructing a training data set adapted to the fashion design, so that the key part morphology of the completed human body is more close to the real human body, and the core demand of garment production is accurately matched.

Generative Face Video Compression (“GFVC”) technology is provided to improve the performance of face video compression. The computational system is configured to: compute a relative difference metric describing the feature differences between frames; and, based on the relative difference metric, determine whether the current frame can be synthesized without entropy coding or should be re-encoded. The computational system is configured to perform two-stage training to stabilize the training of the generative adversarial network (“GAN”) in GFVC.

This invention relates to the field of intelligent clothing design, specifically to a method and system for optimizing down jackets based on generative adversarial networks (GANs) and user feedback. The method includes: a generation module that generates down jacket appearance renderings and structured pattern diagrams based on user body shape parameters and style semantic descriptions using a GAN; a feedback acquisition module that obtains user ratings, comfort evaluations, and local annotation information through virtual or physical try-ons, and collects feedback from professional pattern makers on pattern structure corrections; a feedback learning module that converts the multi-source feedback into sample weighted signals, structural regression signals, and preference prediction signals; a model optimization module that updates the weights of the generation model based on an optimization objective function including adversarial loss, structural constraint loss, and feedback joint loss; and an output module that generates optimized down jacket design results and editable pattern files. This invention can improve down jacket design efficiency and fit rate, and enhance the matching degree between design schemes and user preferences.

This invention relates to the field of digital image security technology, and in particular provides a robust feature embedding image watermarking method based on noise-aware guidance. The method includes acquiring a carrier image dataset, constructing a robust feature embedding image watermarking network based on noise-aware guidance, the robust feature embedding image watermarking network including a watermark embedding architecture and a watermark extraction architecture; inputting the carrier image into the watermark embedding architecture to obtain a watermarked image; inputting the watermarked image into the watermark extraction architecture to obtain watermark information; optimizing network parameters, and simultaneously using a discriminator as an adversarial network, utilizing adversarial loss to ensure that the carrier image and the watermarked image are similar. This method improves robustness against attacks while generating high-quality watermarked images.

This application belongs to the field of computer technology, specifically relating to a deep learning-based method for big data quality assessment and repair. The method includes: collecting multi-source heterogeneous data and performing semantic alignment and integration; extracting features using a deep network with a self-attention mechanism and identifying anomalies based on reconstruction errors; constructing a multi-dimensional indicator system for quantitative quality assessment; performing intelligent repair and semantic verification of the abnormal data based on generative adversarial networks and bidirectional long short-term memory networks; verifying the repair effect and updating the governance strategy. This application, through the above scheme, achieves an automated closed loop from anomaly detection to repair, significantly improving the detection rate of hidden errors and the consistency of repair, while ensuring processing efficiency and data privacy and security.

The application discloses a kind of urban vegetation carbon storage remote sensing estimation methods of mixed Swin Transformer conditional generation adversarial network, including remote sensing image data preprocessing;CNN local feature is constructed and CNN local feature branch is extracted, and Swin Transformer global feature is constructed and Swin Transformer global feature branch is extracted;Multi-scale discriminator is constructed and three-stage training strategy is used, and the final carbon storage prediction result and uncertainty are output.The data input of the application uses open source remote sensing image, improves the estimation accuracy by hybrid network architecture, simultaneously, the uncertainty information of estimated is output synchronously by the probability modeling module built in the model, fills the short board that only " single point numerical result " is provided in prior art, improves the reliability and practicality of model, provides support for urban ecological planning and carbon sink transaction.

The application discloses a multi-working-condition industrial process fault diagnosis method based on a dynamic adaptive domain adversarial network, and comprises the following steps: standardizing process data and taking a sliding window to obtain dynamic data; training a dynamic adaptive domain adversarial network by using labeled dynamic data of a source working condition and unlabeled dynamic data of a target working condition; the dynamic adaptive domain adversarial network comprises a feature extractor, a label classifier, a global domain discriminator and a local domain discriminator; the global domain discriminator aligns the marginal distribution of the source working condition and the target working condition; the local domain discriminator aligns the conditional distribution of the source working condition and the target working condition; a kind of learnable parameter is proposed to adaptively evaluate the relative importance of the marginal distribution and the conditional distribution, so as to better extract domain-invariant features; a kind of common center loss is proposed to improve the inter-class separability and intra-class compactness of data, and further improve the fault diagnosis precision; and the label classifier is used to predict the fault category of a sample. The application establishes an accurate fault diagnosis model for the target working condition.

The invention discloses an image generation method and the like, which are used for generating a high-quality image. The method comprises the following steps: acquiring a target vector; respectively inputting the target vector into a first generator and a second generator to correspondingly generate a first sub-graph and a second sub-graph, the first generator being obtained by training an initially configured first generative adversarial network GAN by a server according to a low-frequency image and a first random noise variable satisfying normal distribution, and the second generator being obtained by training an initially configured second generative adversarial network GAN according to a second random noise variable satisfying normal distribution; the second generator is obtained by training an initially configured second generative adversarial network (GAN) by the server according to the high-frequency image and a second random noise variable satisfying normal distribution, and the frequency of the low-frequency image is lower than that of the high-frequency image; and synthesizing the first sub-image and the second sub-image to obtain a target image.

The application discloses a power grid oscillation risk scene sample generation method, system, device and medium, and particularly relates to the technical field of oscillation scene generation, and the technical points are as follows: a time-frequency fusion generative adversarial network is constructed, the time-frequency fusion generative adversarial network comprises at least one generator and at least one discriminator; wherein the generator comprises a time-frequency decomposition module, a time domain generation module, a frequency domain generation module and a U-Net module; a random noise vector is input into the time-frequency decomposition module to obtain a time domain signal and a frequency domain signal; the time domain signal is input into the time domain generation module, and the frequency domain signal is input into the frequency domain generation module to generate a time domain waveform and a frequency domain feature respectively; the time domain waveform and the frequency domain feature are input into the U-Net module to generate a time-frequency fusion power grid oscillation signal; the generated power grid oscillation signal and an original real oscillation signal are input into the discriminator, and the power grid oscillation signal is iteratively optimized and trained in combination with a loss function to obtain a power grid oscillation scene sample.

This application discloses a data generation method for the field of artificial intelligence technology. The method generates candidate data by inputting random noise into a data generation model within a target generative adversarial network (PGN) built on a semi-supervised learning model based on target business sample data. This automates the data generation process, saving manpower and time costs. The target GAN can learn using a small amount of labeled data and a large amount of unlabeled data, thereby better capturing the inherent characteristics and distribution patterns of the target business data. Candidate data generated in this way is more likely to reflect the actual situation of the target business and has higher authenticity and representativeness. After subsequent screening of target data, this data can be more accurately used to train the prediction model corresponding to the target business, thereby improving the accuracy and reliability of applications such as risk assessment.

The application discloses a highway pavement disease detection method, system and equipment, and relates to the technical field of highway pavement image processing. The technical scheme points are: obtaining a to-be-detected pavement image of a highway; inputting the to-be-detected pavement image into a pre-trained disease area segmentation model for segmentation and extraction to obtain a disease area image of the highway pavement; wherein, a first discriminator and a first generator are used to form a first generative adversarial network, the first generative adversarial network is trained, and when the number of training reaches a maximum number, a disease detection model is obtained; the disease area image is input into the pre-trained disease detection model for detection to obtain a disease detection result of the highway pavement; wherein, a YOLOv8 neural network is trained, and when a first loss function converges, the disease detection model is obtained. The problems that the prior art cannot capture detailed information and has false and missed detection of diseases when detecting pavement diseases are solved.

This invention proposes an intelligent error correction method and system for ethnic vocal music dialect prosody. The method includes: collecting multimodal data, extracting features to form a training dataset, constructing and updating a dynamic dialect prosody map, and performing cross-modal comparative analysis to obtain a joint representation of cross-modal error features. A neural vocoder direct-connect correction model is constructed, inputting the joint representation to generate a corrected Mel spectrum. A cross-modal generative adversarial network is constructed, including a generator and three discriminators. The generator generates a joint latent representation based on high-fidelity corrected audio and three features. The three discriminators respectively judge the naturalness of the audio, the compliance of the text tone, and the rhythmic coordination of the musical score. This invention solves the problem of manual error correction by constructing a prosody map, overcomes the bottleneck of manual detection by utilizing cross-modal comparative analysis, improves the generalization ability of the model by combining meta-learning algorithms, and optimizes the correction results with the help of adversarial networks, thereby reducing manual costs and improving the error correction efficiency of ethnic vocal music works.

The application discloses a kind of track vehicle running gear anomaly detection methods based on symbolic regression and generative adversarial network, comprising the following steps: collecting the sensing data of track vehicle running gear, and save in database;Establish structural learning model and generative adversarial network model, the result of structural learning model and generative adversarial network model is superimposed as health baseline data;Obtain the sensing data of track vehicle running gear in target period as real-time monitoring data, calculate real-time deviation according to real-time monitoring data and corresponding health baseline data;Calculate the average error of all real-time deviations in target period, if average error is greater than preset alarm threshold, it is judged to appear abnormality.The application can accurately analyze the mechanism of the structural characteristics of track vehicle running gear, realize the dynamic real-time tracking of anomaly detection.

This invention discloses a method and system for monitoring exercise fatigue based on multimodal biometric recognition, relating to the field of exercise health monitoring technology. The method includes: simultaneously acquiring surface electromyography (EMG), muscle thickness, and inertial data; decomposing EMG signals to reconstruct pure signals and artifacts, and generating reliability weights based on inertia; extracting electrophysiological features from EMG, constructing hysteresis loops from thickness and angle to extract mechanical features, and extracting emotional features from inertia; inputting the three types of features into an adversarial network to decouple and output pure fatigue features; inputting the data into a hidden Markov model to decode the current state and predict the next cycle; fusing the prediction vector, fatigue features, and weights, generating fused features via a graph network and self-attention, outputting a level and labeling the type. This invention solves the problems of dynamic changes in signal quality leading to fused weight mismatch, difficulty in distinguishing fatigue sources, lack of evolutionary prediction, and low cross-individual adaptation efficiency, achieving high-precision and interpretable real-time exercise fatigue monitoring.

The application discloses a three-dimensional seismic exploration prediction analysis method and system, and relates to the field of three-dimensional seismic exploration.The method comprises data preparation and preprocessing; construction of interpretation and establishment of a geological framework; well-seismic joint analysis and construction of a prior model; construction of a prior probability three-dimensional geological model through Markov chain simulation of lithofacies, Monte Carlo sampling to give elastic parameters, and generation of an adversarial network to expand pseudo-well samples; multi-source information fusion and probabilistic reservoir prediction; extraction of multi-evidence volumes and determination of weights; multi-target threshold decision and delineation of a favorable zone; high-precision target appearance and well site deployment.The method realizes logical judgment before appearance through prior probability modeling and multi-evidence weighted fusion, outputs a three-dimensional probability volume to quantify uncertainty, objectively delineates a favorable zone through multi-target decision, significantly improves reservoir prediction accuracy and micro-geological body recognition capability, and reduces exploration risk.

This invention provides a traffic flow prediction method and system based on elastically regularized generative adversarial networks (GANs) and convolutional neural networks (CNNs), belonging to the field of traffic flow prediction and management technology. The method involves acquiring traffic flow data at the current time step; processing this data using a pre-trained traffic flow prediction model to obtain predicted traffic flow data for multiple future time steps. By introducing an elastically regularized GAN, this invention can automatically complete missing data without manual interpolation or expert rule design, improving the robustness and applicability of traffic flow data recovery; it effectively captures the spatial structural features of traffic flow, achieving multi-scale spatiotemporal feature extraction, thus taking into account both local and global dependencies and avoiding information loss; and it incorporates external environmental and social factors into the prediction model, significantly improving prediction accuracy and generalization ability under abnormal weather conditions or special travel peaks.

This invention provides a method and system for enhancing server load data using generative adversarial networks (GANs), relating to the field of server performance management technology. First, a multi-dimensional raw server load data set collected in a time-series manner is acquired. This data is then decomposed into time-series evolution patterns to obtain sets of trend evolution and periodic fluctuation components. Next, a GAN is invoked to perform distribution adversarial fitting on these components, generating a composite set of newly added load trend evolution and a composite set of newly added load periodic fluctuations. Then, the components are reconstructed and combined to obtain an enhanced load data set. Finally, based on this set, time-series overlay and expansion are performed to generate a target load data set for expanding the server load analysis sample library, providing sufficient high-quality data for server load analysis.

This application provides a method for constructing a container door anomaly detection model and an anomaly detection method and device, which are applied in the field of port automated visual monitoring technology. After preprocessing real anomaly samples, virtual anomaly samples are generated using a Wasserstein generative adversarial network to expand the training dataset, effectively solving the problem of model training difficulties caused by the scarcity of anomaly samples. A deep detection network with a joint attention enhancement mechanism is constructed, and a loss function based on Wasserstein distance is used as the regression loss during the training process, thereby obtaining a detection model that can accurately detect container door anomalies, significantly improving the detection accuracy of the model in complex backgrounds.

This invention discloses a multi-timescale market bidding method and system for energy storage systems considering dynamic opportunity costs, relating to the fields of electricity market and energy storage optimization technology. Through a dynamic opportunity cost quantification model, the potential impact of current decisions on future returns is dynamically quantified as constraints. While avoiding short-sighted decisions, a multi-market time-series coupling mechanism is constructed using the capacity state transfer equation and cross-market power balance constraints. This ensures the physical feasibility and temporal consistency of the dynamic allocation of energy storage capacity across day-ahead, intraday, and real-time markets. Furthermore, a stochastic-robust hybrid optimization framework is proposed, integrating generative adversarial network scenario generation, conditional risk value, and robustness margin to unify the optimization of expected returns and extreme risks. The constraints are transformed into optimization problems using KKT conditions and the Big M method, and solved using the ADMM distributed algorithm, ensuring the model's feasibility and real-time performance in engineering practice.

The application discloses a wearable intelligent collaborative management and control method for power customer service risk, and relates to the technical field of artificial intelligence and data processing. The method comprises the following steps: acquiring real-time interaction of a wearable device and synchronizing power business data; separating style and emotion feature vectors by using an adversarial network; constructing a dynamic reference vector based on the style feature and calculating an emotion deviation degree to generate real-time risk representation; predicting a risk evolution track according to the real-time risk representation; when a trigger condition is met, simulating the effect of a candidate strategy based on counterfactual reasoning and determining a target strategy in combination with a business state. The application is used to solve the problems of high risk identification false alarm rate, intervention feedback lag and difficulty in balancing service benefits and operation safety under the complex environment of power operation.