258 results about "Sample graph" patented technology

The invention relates to the technical field of defect detection, and particularly discloses an electronic component defect detection method, system and device, and the method comprises the steps: collecting a multiband original image set based on multispectral imaging, inhibiting batch material difference and surface reflection through orthogonal complementary space projection, and obtaining a de-noised image set; performing feature extraction on the de-noised image set, and identifying and marking all fragmented defect areas; based on a graph neural network space attention mechanism, establishing an association relationship between fragmented defect areas to form a complete defect feature chain; gray frequency value sequences of the to-be-detected image and the standard sample image are constructed respectively, and a first fitting curve and a second fitting curve are obtained through fitting of a skewed distribution function; and finally, comprehensively judging whether the electronic component has defects or not by combining the defect characteristic chain and the difference characteristics of the two fitting curves. According to the method, interference can be effectively suppressed, fragmented defects can be accurately identified, and the detection accuracy and reliability are remarkably improved through multi-dimensional feature fusion.

A semantic segmentation model training method and apparatus, an electronic device and a storage medium are provided. The semantic segmentation model training method includes: acquiring a sample image, and extracting visual image features corresponding to the sample image by a semantic segmentation model to be trained; processing the sample image to obtain a text image feature corresponding to the sample image, the text image feature being an image feature generated from language description text for the sample image; fusing the visual image features with the text image feature to obtain multimodal features, and performing image segmentation prediction based on the multimodal features to obtain a target loss; and training the semantic segmentation model to be trained based on the target loss to obtain a target semantic segmentation model.

A semantic segmentation model training method and apparatus, an electronic device and a storage medium are provided. The semantic segmentation model training method includes: acquiring a teacher semantic segmentation model that is pre-trained, the teacher semantic segmentation model including a first teacher network and a second teacher network, the first teacher network having structural characteristics of low depth and high width, and the second teacher network having structural characteristics of high depth and low width; processing a sample image based on the teacher semantic segmentation model to obtain a first segmentation map and a second segmentation map; and training a student semantic segmentation model that is lightweight according to the sample image, the first segmentation map and the second segmentation map, so as to obtain a target semantic segmentation model.

The invention is suitable for the technical field of artificial intelligence, and provides an abnormal sample image generation method, electronic equipment and a storage medium, and the method comprises the steps: setting a material parameter library and a scene parameter library, and constructing a paired industrial defect sample data set in combination with three-dimensional geometric models of a plurality of sample workpieces; constructing category text description, defect text description and material category text description of various workpieces, and constructing and training an abnormal sample image generation model in combination with the paired industrial defect sample data set; obtaining a target normal image, a candidate defect area mask image, a target category text description, a target defect text description and a target material category text description of a target category workpiece, and inputting the target normal image, the candidate defect area mask image, the target category text description, the target defect text description and the target material category text description into an abnormal sample image generation model for processing to obtain a target defect image and a target defect area mask image; the training precision and flexibility of the abnormal sample image generation model are improved, and then the efficiency and precision of abnormal sample generation are improved.

The invention relates to the technical field of image generation, and discloses a high-quality human body image generation method and device, and computer equipment. The method comprises the following steps: constructing a training data set, wherein the training data set comprises a sample image of a human body, a text description of a corresponding human body attribute, a human body analysis graph and a plurality of attribute tags; an image generation network is constructed, the image generation network comprises an encoder, a UNet and a decoder which are connected in sequence, the image generation network further comprises a time feature aggregation module and an attribute perception reward module, and the attribute perception reward module is used for predicting a target reward score and a prediction reward score; training an image generation network by minimizing the difference between the target reward score and the predicted reward score; and inputting the text description of the human body image to be processed and the human body analysis graph into the trained image generation network to obtain a target image. By adopting the method, the high-quality human body image with space alignment and consistent attributes can be generated.

The disclosure provides a graph neural network training method, a training system and an abnormal account identification method. The graph neural network training method comprises: obtaining initial graph structure data corresponding to a terminal device; the initial graph structure data obtained by a plurality of distributed training terminals respectively is derived from the same sample graph structure data; the following graph structure data processing stage and graph neural network training stage are executed cyclically until a target neural network meeting the training requirement is obtained: determining a processing time of the current execution graph structure data processing stage according to historical execution data of the historical execution graph structure data processing stage and the historical execution graph neural network training stage; performing graph structure data processing on the initial graph structure data in the graph structure data processing stage according to the processing time to generate target graph structure data; the graph structure data processing comprises data sampling processing and feature extraction processing; and training the target neural network based on the target graph structure data in the graph neural network training stage.

The invention relates to the technical field of intelligent operation and maintenance of an electric power system, in particular to an inspection sample image data enhancement method for an electric power artificial intelligence platform, which is used for solving the problems that in the prior art, history and equipment knowledge cannot be fused to construct a forbidden area, a co-occurrence rule and component association, cross-component defect positions cannot be effectively adjusted, and the detection accuracy is poor. Defect distribution is difficult to accurately control, and physical rationality and engineering credibility are reduced. According to the method, a forbidden area, a co-occurrence rule and component association are constructed by fusing history and equipment knowledge, masks are generated through kernel density estimation to suppress invalid defects, co-occurrence frequencies are counted based on feature vectors, co-occurrence relationships are determined by combining distances and similarities, and the masks, matrixes and graphs are embedded and coded into conditional vectors, so that the non-ineffective defects are suppressed. And zero setting is performed on a forbidden area in the generative network, illegal co-occurrence is filtered, and cross-component defect positions are adjusted, so that defect distribution is accurately controlled, and physical rationality and engineering credibility are enhanced.

The invention discloses an industrial image anomaly detection method based on a multi-agent arrangement heterogeneous algorithm. The method comprises the following steps: S1, constructing an algorithm component library for shielding data representation differences among different architecture algorithms; s2, constructing a multi-modal large model system based on a multi-agent architecture; the multi-agent architecture comprises the following steps: receiving an image by using a visual expert agent, and outputting structured physical metadata and an unstructured visual suggestion text; the planner agent receives the physical metadata and the visual suggestion text, and generates an algorithm configuration tree based on an algorithm component library; the optimizer agent optimizes the algorithm configuration tree in the closed-loop feedback stage; s3, using the algorithm configuration tree to construct a reference model representing good product distribution according to the normal training sample images; s4, performing anomaly detection on the to-be-detected sample image by using the algorithm configuration tree to generate an anomaly judgment result; and S5, optimizing the algorithm configuration tree.

The embodiment of the application provides an image enhancement joint optimization method and system based on downstream task performance, and relates to the technical field of image processing. The method first acquires a training data set and task evaluation data, inputs sample images in the training data set into an image enhancement model to obtain enhanced images, and inputs the enhanced images into a task model to obtain task output results. Then, a target training mode is determined according to the task evaluation data, and a parameter update strategy is set based on the target training mode. Then, according to the parameter update strategy, task metric information is calculated according to the task output results and task annotation information, and the model parameters of the trained model are updated according to the task metric information. The method can form an evaluation and closed-loop optimization mode based on the performance of the downstream task as the core constraint, and by introducing the task metric information under controllable conditions, a task-driven optimization strategy and a stabilization mechanism are used in the training stage, so that the robustness and stability of the image processing process are improved.

The invention discloses an ultrasonic pulmonary artery segmentation method, system and equipment based on anatomical perception and a medium, belongs to pulmonary artery segmentation in an artificial intelligence technology, and aims to solve the technical problems of low accuracy and poor segmentation integrity of pulmonary artery segmentation. Comprising the steps of sample image acquisition, classification network fine adjustment and feature extraction, initial response diagram generation, pseudo label generation, segmentation network construction and first training, dynamic anatomy constraint iteration training and pulmonary artery real-time segmentation. When a pseudo label is generated, constructing an anatomical perception pseudo label generation module, inputting the initial response graph into the anatomical perception pseudo label generation module, and outputting an initial pseudo label by the anatomical perception pseudo label generation module; and during iterative training, constructing a differentiable anatomical constraint generator, and performing iterative training on the segmentation network by using the differentiable anatomical constraint generator. A rough and sparse initial response graph is converted into a complete, continuous and reasonably dissected initial pseudo-tag through a pseudo-tag generation module for dissecting perception.

The invention provides a data security risk assessment method and system. According to the method, a data access request is acquired, a target tracking identifier corresponding to the data access request is generated or received for the data access request based on a target entry service, and inter-service calling related to the data access request is tracked according to the target tracking identifier to acquire calling link data corresponding to the data access request. On the basis of pre-established sensitive field metadata, labeling a field access behavior of target business data involved in calling link data as a corresponding sensitive data access event, constructing a data stream sample graph according to an association relationship between the sensitive data access event and a target tracking identifier, performing risk analysis on the data stream sample graph, and obtaining a sensitive data stream; therefore, the data security risk assessment result corresponding to the single data access request can be output at the request level, and the full-link assessment of the data security risk under the micro-service architecture is realized.

The application provides a multimodal large model security protection method, device and equipment, which comprises the following steps: obtaining a sample image and a sample text, and performing feature fusion on the sample image and the sample text to obtain a multimodal feature; performing a first perturbation operation on the multimodal feature to obtain a first perturbation feature, inputting the first perturbation feature into an initial multimodal large model to obtain a first prediction label; determining a first loss value and a second loss value based on the first prediction label; adjusting the initial multimodal large model based on the first loss value to obtain an intermediate multimodal large model; performing a second perturbation operation on the multimodal feature to obtain a second perturbation feature, inputting the second perturbation feature into the intermediate multimodal large model to obtain a second prediction label, and determining a third loss value based on the second prediction label; and adjusting the initial multimodal large model based on the second loss value and the third loss value to obtain a target multimodal large model. Through the application scheme, the calculation resources are saved, and the training time is reduced.

The invention discloses an ultra-sparse CBCT (cone beam computed tomography) reconstruction method, system and equipment based on an ordinary differential equation, belongs to CBCT reconstruction in the field of artificial intelligence, and aims to solve the technical problem of low quality of CBCT reconstructed images. The method comprises the following steps: acquiring sample data, preprocessing the data, constructing and training a CBCT-CT nonlinear relation reconstruction model, and performing real-time reconstruction; during preprocessing, converting the three-dimensional image volume data into simulated X-ray projection data, and reconstructing the simulated X-ray projection data by adopting an FDK reconstruction algorithm to obtain an FDK-CBCT image; the CBCT-CT nonlinear relation reconstruction model comprises an encoder, a NODE module and a decoder; in the training process, the CBCT-CT nonlinear relation reconstruction model is trained through the obtained CT sample image and the FDK-CBCT image. In the reconstruction model, through continuous evolution of NODE modeling image features, the model can model a continuous evolution mapping process from a sparse low-quality image to a high-quality CT image during training, so that stripe artifacts and structural distortion do not easily exist in the reconstructed image, and the reconstruction quality is high.

This application relates to the field of industrial defect detection technology, specifically to a method for establishing a feature library in defect detection, a defect detection method, and an apparatus, which can, to some extent, solve the problem of redundant data in the feature library of defect detection methods. Positive sample image data is processed through a pre-trained first network model to extract a first initial feature vector, and an initial feature library is established based on the first initial feature vector. A second network model is triggered to fit the first network model, and a second feature vector is extracted from the positive sample image data using the fitted second network model. Based on the degree of matching between the second feature vector and the corresponding first initial feature vector, the frequency level corresponding to the first initial feature vector is determined. A first candidate feature vector (a first initial feature vector with a frequency level greater than the first level or less than the second level) is selected, and a target feature library is established based on the first candidate feature vector, reducing redundant data in the feature library.

The disclosure provides a method for training a graph neural network. The method includes: obtaining initial graph structure data corresponding to the terminal device, initial graph structure data respectively obtained by distributed training terminals being derived from the same sample graph structure data; and performing a graph structure data processing stage and graph neural network training stage cyclically, until a target neural network satisfying a training requirement is obtained: determining a processing opportunity for currently performing a graph structure data processing stage based on historical execution data of historically performing a graph structure data processing stage and a graph neural network training stage; performing, based on the processing opportunity, graph structure data processing on initial graph structure data in the graph structure data processing stage, to generate target graph structure data; and training, based on target graph structure data, the target neural network in the graph neural network training stage.

The invention provides an image processing method and device based on a large model, a medium, electronics and a product, and relates to the technical field of image processing, the method comprises the following steps: obtaining a target image and a target cue word, the target cue word being obtained by performing multiple rounds of iterative optimization on an original cue word according to a target large model and a plurality of optimization samples; wherein the optimization sample comprises a sample image and a specified processing result corresponding to the sample image, in the optimization process, the target large model is used for processing the sample image and the original cue word to obtain a prediction processing result, and an optimization abstract is obtained according to historical data, the prediction processing result and the specified processing result; the original cue word is optimized according to the optimization abstract; and obtaining a target processing result through the target large model according to the target image and the target cue word. Automatic optimization of the cue word can be realized, and the accuracy of the obtained target cue word can be improved, so that a more accurate target processing result is obtained.

The application discloses a feature placement method and system for multi-GPU sampling graph neural network training, uniformly models a sampling graph neural network training process, and obtains hardware topology information; in a pre-sampling process, the number of times of accessing node features is counted as node heat, the node features are sorted according to the node heat, the node features are divided into a plurality of feature blocks according to a single-block capacity, and the feature blocks are divided into hot blocks or cold blocks according to block heat; based on the feature block set obtained by the division and the obtained hardware topology information, a mixed integer programming model is constructed, with the minimum bottleneck link time in a feature extraction stage and the memory load imbalance degree as targets, the mixed integer programming model is solved, and the placement position, access path and link flow of the globally jointly optimized hot block are obtained. The application can obtain a high-quality feature placement scheme under a given hardware constraint, minimize the extraction time, and improve the resource utilization rate of a multi-GPU system.

The application discloses a method, device and equipment for detecting illegal advertising boards based on deep learning and a medium, wherein the method comprises the following steps: constructing a sample set and a test set; inputting sample images in the sample set into a YOLOv7 model for detection, and outputting detected targets; performing character recognition on the detected targets, comparing all words in the recognized text with a pre-created word library, searching for words with the highest similarity in the word library as keywords and corresponding probability values; using a loss function to optimize the model training process; using sample images in the test set to test the optimized YOLOv7 model, and detecting illegal advertising boards for to-be-detected images after the test is completed. The application enhances the feature expression capability, uses keywords obtained through character recognition to assist in training, is beneficial to distinguishing illegal advertising boards from regular advertising boards, and improves the detection accuracy.

The application relates to a remote sensing image generation online map method, device, equipment and medium, a plurality of different transformed sample images are obtained by performing geometric transformation on remote sensing sample images in each group of sample image pairs, the plurality of transformed sample images are added to the corresponding sample image pairs to obtain a training data set, each training data set is used to train an online map generation model to obtain a trained online map generation model, an encoder in a perception image compression network is used to map remote sensing sample images and corresponding transformed sample images from a pixel space to a feature space, a forward diffusion and reverse denoising are performed on the feature space by a denoising diffusion bridge network, and then an output of the denoising diffusion bridge network is mapped from the feature space to the pixel space by a decoder in the perception image compression network, and a real-time remote sensing image is input into the trained online map generation model to obtain a real-time network map. By adopting the method, a map with clearer boundaries and brighter colors can be generated in real time.

The present disclosure provides a large model-based image processing method, device, medium, electronic and product, relating to the technical field of image processing, which comprises: acquiring a target image and a target prompt word, the target prompt word being obtained by performing multi-round iterative optimization on an original prompt word according to a target large model and multiple optimization samples; wherein the optimization sample comprises a sample image and a specified processing result corresponding to the sample image; in the optimization process, the target large model is used to process the sample image and the original prompt word to obtain a predicted processing result, and an optimization summary is obtained according to historical data, the predicted processing result and the specified processing result, and the original prompt word is optimized according to the optimization summary; and a target processing result is obtained through the target large model according to the target image and the target prompt word. The automatic optimization of the prompt word can be realized, the accuracy of the obtained target prompt word can be improved, and a more accurate target processing result can be obtained.

The invention provides a traffic scene-oriented zero sample target detection method based on comparative learning, which comprises the following steps of: constructing an enhanced label semantic description for a traffic target category, encoding the label semantic description by using a text encoder, and obtaining a target feature and functional attribute information of the target, wherein the label semantic description comprises visual features and functional attribute information of the target; therefore, semantic features are generated; extracting sample image features of the sample image through an image encoder; carrying out hierarchical comparison learning on the sample image features and the semantic features so as to align the sample image features with the semantic features; according to the method, query image features of a query image are extracted through an image encoder, and detection and semantic matching of a query image target are realized based on the similarity between the query image features and sample image features and the similarity between the sample image features and semantic features. The purpose of improving the accuracy and robustness of traffic target detection is achieved.

This application discloses an automated rendering optimization method based on visual UI. The method provides optimization for rendering images containing text content. To obtain a rendered image that achieves the desired effect in terms of graphic content, text content, and the fusion between the two, the fusion degree between the two is quantified during sample construction. This helps to convey information while expressing the aesthetic appeal of the image and restoring the designer's intention. Furthermore, the sample labels also indicate the position of the text content in the sample graph, allowing the image processing model to understand the relationship between the boundary and fusion degree between graphic and text content, thus better locating the text content. After the image processing model converges, it possesses the ability to locate the text content and distinguish it from the graphic content. This application provides the means to identify whether the rendering effect of an image needs further optimization.

The application relates to the technical field of computer vision, and particularly discloses an image classification model construction and training method based on difficult-easy sample correlation learning, which comprises the following steps: processing sample images in a training set and constructing an image classification model; training the image classification model by using all the processed sample images in the training set to obtain a pre-training image classification model; performing inference on all the processed sample images in the training set by using the pre-training image classification model, and dividing the training set into a difficult sample set and an easy sample set; and optimizing and training the pre-training image classification model based on a correlation learning mechanism of the difficult sample set and the easy sample set, updating the difficult sample set and the easy sample set after each round of optimization and training, and obtaining a final image classification model after multiple rounds of optimization and training. The application can reduce the bias caused by class distribution imbalance, enhance the capture learning of difficult sample features by the model, and improve the precision of image classification and recognition.

The invention provides a model training method and device, an image generation method and device, equipment and a storage medium, and relates to the technical field of computers. In order to efficiently distill the semantic feature understanding ability of a teacher model to a student model and improve the reproduction ability of the student model to the semantic features of a key area, the model training method provided by the invention comprises the following steps: inputting a sample image into a pre-training model to obtain a first feature map output by a first intermediate layer of the pre-training model; inputting the sample image into a to-be-trained model to obtain a second feature map output by a second intermediate layer of the to-be-trained model; determining a first weight corresponding to each image area of the sample image according to a first difference between the first feature map and the second feature map; and training the to-be-trained model based on the first weight by taking learning of the first intermediate layer by the second intermediate layer as a target to obtain the trained student model, thereby improving the processing precision and effect of the trained student model.

The invention discloses a model training method and device, a target detection method and device, electronic equipment and a storage medium. The training method comprises the following steps: inputting a sample image and sample point cloud data collected for a sample region into a teacher model for target detection to obtain first feature information and a first detection result; inputting the sample image into a student model for target detection to obtain second feature information and a second detection result; determining target loss according to at least two of the first distillation loss, the second distillation loss and the third distillation loss; and at least adjusting parameters of the student model according to the target loss until a training ending condition is met, and taking the student detection model meeting the training ending condition as a target detection model. According to the method provided by the invention, the accuracy of target detection is effectively improved.

The invention discloses a multi-GPU sampling type graph neural network training-oriented feature placement method and system, and the method comprises the steps: carrying out the unified modeling of a sampling type graph neural network training process, and obtaining hardware topology information; in the pre-sampling process, counting the access times of the node features as node popularity, sorting the node features according to the node popularity, dividing the node features into a plurality of feature blocks according to single block capacity, and dividing the feature blocks into hot blocks or cold blocks according to block popularity; on the basis of the feature block set obtained through division and the obtained hardware topological information, a mixed integer programming model with the purpose of minimizing bottleneck link time and video memory load imbalance in the feature extraction stage is constructed, and the mixed integer programming model is solved; and obtaining the placement position, the access path and the link flow of the hot block subjected to global joint optimization. According to the method, a high-quality feature placement scheme can be obtained under given hardware constraints, so that the extraction time is minimized, and the resource utilization rate of a multi-GPU system is improved.

The embodiment provides a crack identification model training method and device, electronic equipment and a storage medium. The input data set of the model can be generated by using a random cosine function conforming to the formation crack characteristics, sample expansion in a small sample condition is realized, data acquisition work is simplified, cracks in a sample image can be automatically labeled, human and time costs are saved, a Cluster NMS is used to replace an NMS of a yolo v5s original model to perform a redundant frame deletion operation, a weight and a center point distance penalty term are added to complete non-maximum suppression, the model can more quickly and accurately obtain an optimal frame, and therefore, target detection effect is improved, which is of great significance for crack reservoir exploration.

The application provides a matching method, device and system of a house type drawing and an effect drawing, and a storage medium. The method comprises the following steps: obtaining a 2D house type drawing, and converting each independent space region in the 2D house type drawing into house type vector data; extracting house feature information in the house type vector data, and constructing a feature sample drawing based on the house feature information; comparing the feature sample drawing with a feature index drawing; extracting a feature index drawing with the highest matching degree, and obtaining an effect drawing corresponding to the feature index drawing. The method provided by the application realizes the matching association and retrieval between the 2D house type drawing and the effect drawing, enables the 2D house type drawing and the effect drawing to establish bridge data related to each other according to the human brain thought, avoids the defects of manual page-by-page searching and looking up, improves the matching searching efficiency and accuracy, and provides support for pushing services, data statistics, garbage data elimination and the like.

The present disclosure provides a training method and device, computer equipment and a storage medium, wherein the method comprises: obtaining a plurality of sample images; the sample images comprise a target object; for each sample image, processing the sample image by using a target neural network to be trained to obtain a plurality of category prediction information corresponding to the target object; wherein different category prediction information is respectively output by different prediction branches in the target neural network; the initial network parameters corresponding to different prediction branches are different; and based on the plurality of category prediction information, determining the target prediction branch corresponding to each sample image from the plurality of prediction branches; based on the plurality of sample images and the target prediction branch corresponding to each sample image in the plurality of sample images, training the target neural network until a preset training stop condition is met, and obtaining the trained target neural network.

The application provides a few-shot graph-level anomaly detection method based on structure perception prompt, comprising: pre-training a detection model to obtain a pre-trained detection model; fine-tuning the detection model by using labeled graphs to obtain a fine-tuned detection model, wherein the labeled graphs comprise labeled normal graphs and labeled abnormal graphs. The model pre-training comprises: constructing an enhanced graph set and a sampled graph set for an original graph set; constructing a graph-graph structure for the original graph set, the enhanced graph set and the sampled graph set; performing message propagation on the graph-level representation on the graph-graph structure to generate a graph embedding matrix; and performing joint optimization of node-graph contrast learning and graph-graph contrast learning on the detection model by using the graph-level representation and the graph embedding matrix to obtain the pre-trained detection model. The application effectively improves the detection accuracy and generalization ability of the model.