100 results about "Image classifier" patented technology

A network security system that classifies webpages and uses the classification of the webpages to enforce relevant security policies is disclosed. To classify a webpage, a screenshot of at least a portion of the webpage is captured. An embedding engine generates a subject image embedding of the screenshot. The subject image embedding is classified by an image classifier that includes an index of training image embeddings each having a label of their classification and an associated approximate nearest neighbors model trained to identify the label of the closest training image embedding to the subject image embedding and a score representing their similarity. The webpage is classified based at least in part on the score and the label from the image classifier. The network security system applies security policies to requests from client devices that identify the webpage as its destination based at least in part on the webpage classification.

Attention guided images for updating predictions made by an image classifier model are generated by extracting feature maps from a last convolutional layer or other layers of the image classifier model, where the image classifier model generates the feature maps in response to being provided as input an image and computing as output a prediction of a diagnosis. A concept weighted feature map is generated based on the extracted feature maps and coefficients of a concept class associated with an image class of the image. A weighted average feature map is generated by implementing channel wise pooling by averaging across channels for the concept weighted feature map. An attention guided image is generated by implementing element wise multiplication of the weighted average feature map with the image. An updated prediction of the diagnosis is generated by the image classifier model based on the attention guided image.

The invention belongs to the field of deep neural networks and image classification, and particularly relates to an image classification adversarial training improvement method based on boundary sample enhancement. Comprising the following steps: step 1, preparing an image sample set; 2, designing a GAN network architecture suitable for boundary sample generation; step 3, training a GAN network suitable for boundary sample generation; and step 4, image classifier adversarial training participated by the boundary samples. The method can effectively restrain the stability of the decision boundary of the image classification model, enables the model to carry out finer feature learning in a critical region, and prevents the boundary from excessively shifting due to adversarial disturbance. And meanwhile, the model can still maintain stable and reliable feature expression when being disturbed. According to the method, the robustness of the model is improved, the damage to the precision of a natural sample is reduced, and better accuracy-robustness balance and stronger generalization ability are realized.

The invention relates to the field of industrial network security, and particularly discloses a picture network intrusion detection method based on feature selection and data balance, which comprises the following steps of: firstly, acquiring an intrusion detection data set containing non-numeric data and numeric data, performing one-hot coding on the non-numeric data, performing maximum and minimum normalization on the numeric data, and performing one-hot coding on the numeric data; dividing a training set and a test set; performing feature selection on the training set by using a mixed feature selection strategy of a ReliefF filtering method and a Boruta packaging method, and performing resampling processing on the training set by using an SMOTE-ENN mixed sampling method; then, converting the resampled training set and the original test set into grayscale images, and performing training by using an OfficientNet image classifier to obtain a trained model; and finally, the model is used for carrying out classification prediction on gray level images of a test set. The problems of data redundancy, dimension disasters and class imbalance in an IIOT environment are effectively solved, and the accuracy and stability of an intrusion detection system are remarkably improved.

A technique for a UAV includes: acquiring an aerial image of an area below a UAV that includes one or more instances of an object; analyzing the aerial image with an image classifier to classify select pixels of the aerial image as being keypoint pixels associated with keypoints of the object; grouping the keypoint pixels into one or more groups each associated with one of the instances of the object, wherein first keypoint pixels of the keypoint pixels are grouped into a first group of the one or more groups associated with a first instance of the one or more instances of the object; generating an estimate of a relative position of the UAV to the first instance of the object based at least upon a machine vision analysis of the first keypoint pixels; and navigating the UAV into alignment with the first instance based upon the estimate.

An example operation may include at least one of determining, by a transformer encoder trained on annotated image-text data, first latents for a first dataset stored in a memory, and second latents for a second dataset stored in the memory, generating a similarity matrix based on comparisons between the first latents and the second latents, constructing a graph comprising nodes corresponding to the first latents and edges based on pairwise similarity exceeding a threshold, identifying connected components in the graph and selecting, from each component, at least one latent having a highest score from a classifier trained to approximate divergence between the first dataset and the second dataset, forming a reduced dataset comprising the at least one latent, providing the reduced dataset to a model training module, and training an image classifier using the reduced dataset and the second dataset.

The invention discloses a training method and device of an image classifier, relates to the field of data processing, and is used for solving the problem of low accuracy of model recognition image classification in an industrial scene. The training method comprises the following steps: for a first training stage, inputting any first sample data into an initial model to obtain a first classification result output by a main classifier of the initial model and a second classification result output by an auxiliary classifier of the initial model, jointly adjusting parameters of a feature extractor, a main classifier and an auxiliary classifier in the initial model through the first classification result and the second classification result; for a second training stage, inputting any second sample data into the initial classifier obtained in the previous training stage to obtain a third classification result output by the main classifier and a fourth classification result output by the auxiliary classifier; and adjusting parameters of the feature extractor according to the third classification result and the fourth classification result. Through the scheme, the accuracy of the image classifier can be improved in an industrial scene.

The application discloses an image classification method and device. The method comprises the following steps: performing semantic-guided attention weight processing on an image to be classified, extracting attribute visual primitive features and object visual primitive features, and processing the features by using an image classifier to obtain a target classification result corresponding to the image. The image classifier comprises fine-grained attribute prototypes and fine-grained object prototypes. Each fine-grained attribute prototype is calculated based on attribute visual primitive features extracted from an image sample combined with corresponding attribute semantic features and object semantic features, and the fine-grained object prototypes are calculated in the same way. Based on this, each attribute semantic feature or object semantic feature has a plurality of fine-grained attribute prototypes or fine-grained object prototypes corresponding thereto, the visual diversity of each attribute semantic feature and object semantic feature is improved, and the accuracy of image classification is improved.

Computer-implemented method for running an image classifier with an artificial neural network, characterized in that an output quantity (608) of the artificial neural network is determined, wherein the output quantity is determined (606) from input data for the artificial neural network and from an activation function of the artificial neural network and characterizes a classification of the input data, wherein the activation function is at least piecewise continuously derivable and monotonically increasing, wherein the activation function has at least three fixed points, wherein the number of fixed points of the activation function is either finite and odd or countable and discrete, and wherein the derivative of the activation function is alternately monotonically increasing and monotonically decreasing in regions between adjacent fixed points.

The application discloses a glioma IDH mutation state noninvasive prediction system based on a machine learning algorithm, relates to the technical field of medical image processing, and comprises a data acquisition module, a data processing module, a feature extraction module, a single nuclear magnetic sequence image classifier training module, a multi-nuclear magnetic sequence image classifier training module, a model evaluation module and a prediction module; the system acts on conventional nuclear magnetic sequence image data and FW sequence image data, FA sequence image data and MD sequence image data obtained by processing DTI sequence image data, and combines a machine learning algorithm to construct a prediction model, so that the extraction and analysis of features are more reliable and accurate, the limitation of DWI sequence image data in analyzing an IDH mutation state is solved, and the accuracy of the established prediction model is ensured.

The present invention belongs to the field of brain-computer interface technology application technology, and discloses a brain-computer information fusion classification method and system for shared subspace learning, wherein the brain-computer information fusion classification method includes a training phase and an inference phase; wherein the training phase utilizes paired images and brain response data, optimizes the shared subspace model parameters of images and brain responses through a comparative learning strategy of positive and negative sample sampling, and trains an image classifier; the inference phase extracts image features for classification, and achieves the application goal of the entire brain-computer information fusion classification system. The brain-computer information fusion classification system for shared subspace learning of the present invention can train shared subspaces end-to-end, achieve efficient transfer of brain cognitive information, and improve the performance of image classification tasks in complex open scenarios; through the application of "brain out of the loop", it improves efficiency and stability in real-world applications, and has broad application prospects under the new paradigm of brain-computer information collaboration.

The present invention belongs to the technical field of image processing, and discloses a method, a system, a medium, a device and a terminal for detecting a strong light area in an image, including: after preprocessing the image acquired by a camera to complete pixel gray value normalization, performing block processing on the image, calculating the image gray value features, and performing dimensionality reduction processing on the image features by using the PCA method; inputting the image features into an image classifier model to determine the types of strong light and non-strong light in the image; for the image classified as the strong light type, using the gray feature information of the image block to determine the position and size of the strong light area in the image, and completing the detection and positioning of the strong light area in the image. The present invention combines a machine learning image classification algorithm with a traditional image processing method, improves the detection efficiency and accuracy of strong light images, and completes the positioning of the strong light area in the image. The present invention uses the PCA principal component analysis method to optimize the image features and simplifies the operation process of the model.

A computer-implemented method for ascertaining at least one systematic error during the classification of images into at least one image category by a classification algorithm. The method includes: providing a list of image features and respective value indications for each of the image features; applying combinatorial testing to ascertain a predetermined sequence of test cases, which each include a subcombination of the image features and / or value indications included in the list; generating at least one image file using a text-into-image generation algorithm, the file including a synthetic image assigned to the predetermined image category and has the image feature and the value indication; classifying the generated, synthetic image by means of the classification algorithm into at least one of the plurality of image categories; and ascertaining the misclassification value by comparing the image category classified by the classification algorithm with the predetermined image category.

This invention provides a final layer training method and apparatus for a learning model that utilizes text as a substitute for images, reduces annotation costs for data group information where class sets and pseudo-attribute sets are matched, does not require the collection of separate image datasets balanced between data groups, and mitigates bias in image classifiers. [Solution] The final layer learning method of a learning model, performed by the final layer learning device of a learning model, includes the step of inputting training text into the second learning model and training the final layer of the first learning model, based on a projection model that connects the first embedding space of the first learning model and the second embedding space of the second learning model.

A universal model-based modeling and detection method and system for defect detection in panel production lines, a device, and a storage medium. The method comprises: step 1, collecting panel production line sample data and performing defect annotation to obtain a sample data set; step 2, using the data set for modeling and training to obtain a localization model; step 3, on the basis of the data set, cropping images each containing a defect and classifying the images to form defect image data sets; step 4, using the defect image data sets for modeling and training to obtain an image classifier, and cascading the image classifier with the localization model; and step 5, on the basis of factory attributes and manufacturing process attributes, mapping defects to specific defect names by means of the image classifier and the localization model. In the method, accurate defect name mapping can be performed according to specific production environments and requirements. The overall solution reduces communication and checking time, increases sales, and also improves the applicability and accuracy of detection results.

The present invention relates to a computer-implemented method 1000 for identifying patterns (23) that correlate with correct image classifications (52) and incorrect image classifications (52) using a scene graph (12). [Solution] For a set of images (11), this method obtains a scene graph (12) from the images, classifies the images using a pre-trained classifier (5), groups the scene graphs (12') according to whether the classification is correct, extracts representative subgraphs within each group, thereby revealing patterns associated with the classification (52).

A road safety device, comprising: projecting a light pattern 308 on a road; a camera capturing images of the pattern and an overtaking vehicle 304; a classifier model; a processor to analyse distortio

A network security system that classifies webpages and uses the classification of the webpages to enforce relevant security policies is disclosed. To classify a webpage, a screenshot of at least a portion of the webpage is captured. An embedding engine generates a subject image embedding of the screenshot. The subject image embedding is classified by an image classifier that includes an index of training image embeddings each having a label of their classification and an associated approximate nearest neighbors model trained to identify the label of the closest training image embedding to the subject image embedding and a score representing their similarity. The webpage is classified based at least in part on the score and the label from the image classifier. The network security system applies security policies to requests from client devices that identify the webpage as its destination based at least in part on the webpage classification.

The invention relates to a system (200) for detecting one or more abnormalities in an x-ray image using an image classifier and one or more feature extractors. An abnormality is indicative of a pathology, a disease or a clinical finding present in the x-ray image. The feature extractors extract respective image quality features from the x-ray image indicative of a suitability of the x-ray image for detection of the abnormalities. The one or more feature extractors are applied to the x-ray image to determine the respective image quality features for the x-ray image. The image classifier is applied to the x-ray image to determine the classification scores for the one or more abnormalities. The image classifier has been trained to use the determined image quality features to determine said classification scores. A classification result is output based on the determined classification scores.

The application discloses a boundary sample data enhancement method and device for knowledge distillation and a computer storage medium. The method comprises the following steps: before knowledge distillation is performed, the output of a teacher model is used to modify samples in each original data set along the decision boundary of the teacher model step by step, and a plurality of boundary samples suitable for knowledge distillation are expanded. In each iteration, the original sample or each sample modified in the last iteration is used as a basic sample, the approximate tangent plane of the decision boundary near the sample is calculated by using the output of the teacher model, and the sample is modified along multiple directions on the tangent plane; then, the modified sample is modified to be located near the boundary; finally, a plurality of samples farthest from other basic samples are selected as the result of the modification in the round and the basic samples for the next iteration. The application can meet the demand for data enhancement in current image classifier knowledge distillation.

A method for quantitatively rating a trained generator that generates, from an input vector from a predetermined input distribution in connection with a target segmentation map, an image whose semantic content is in line with this target segmentation map. The method includes: drawing at least one input vector from the predetermined input distribution; drawing at least one list of classes from a predetermined class distribution; determining a target segmentation map, which assigns classes from this list of classes to the pixels of the image to be generated; generating, by means of the generator, an image from the input vector and the target segmentation map; determining a semantic segmentation map of the image by means of an image classifier; determining, using a predetermined metric, the degree of matching between this semantic segmentation map and the target segmentation map; determining, using this degree of matching, the quantitative rating of the generator.

A method of predicting user throughput in a network using machine learning (ML) comprises: obtaining a satellite or aerial image of an area; providing it to an image classifier to output recognised an

Systems and methods are provided for image classification using a histogram of oriented gradients (HoG) in combination with a trainer. By first establishing a bitmap that identifies a subset of pixels in a HoG window as including relevant foreground information, and limiting the HoG computation and comparison process to only include pixels in the bitmap, the efficiency of the process is greatly increased.

The application provides a classifier training method, an image analysis method, a device, equipment and a storage medium, wherein the classifier training method is used for training of a tongue image classifier, and the classifier training method comprises the following steps: obtaining a training sample set, the training sample set comprising a first training sample set and a second training sample set, the first training sample set comprising a first preset number of clear tongue images, and the second training sample set comprising a second preset number of blurred tongue images; performing image preprocessing on the clear tongue images and the blurred tongue images respectively to obtain a first image and a second image, the pixels of the first image following a Gaussian distribution, and the pixels of the second image not following the Gaussian distribution; and calculating generalized Gaussian distribution parameters, asymmetric generalized Gaussian distribution parameters and motion blur parameters of the first image and the second image to obtain a training data set. The tongue image classifier obtained through the classifier training method can classify clear tongue images and blurred tongue images.

The invention discloses an infrared image conversion method suitable for a visible light target recognition model, and the method comprises the steps: training an image processing network model which is used for converting an infrared image into a visible light image in the following mode: inputting infrared image sample data into the image processing network model, and obtaining a converted image sample, and inputting the converted image sample into a feature extraction network in a trained visible light target recognition model, carrying out target feature extraction, and inputting the extracted target feature sample into a trained first target classifier and a trained image classifier respectively, and taking the image classification loss function value of the image classifier and the target classification loss function value of the first target classifier as supervision loss function values, and adjusting model parameters of the image processing network model until the supervision loss function values reach the expectation. The method is suitable for the infrared image under the condition that the visible light target recognition model is kept unchanged.

Classifying an object in images of an environment by: receiving an image and a hyperspectral image of the environment; classifying an object in the image using a trained image classifier; labelling each pixel of the hyperspectral image with a material based on its hyperspectral signature to construct a semantic materials map; identifying the classified object in the semantic materials map; verifying the classified object as valid or invalid based on whether the material of the object in the semantic materials map matches an expected material for the classified object. The image may be an RGB image. The object may be military such as an armoured vehicle or tank. Hyperspectral curves for pixels may be constructed by dividing sensed light into frequency bands which are grouped into frequency ranges including ultraviolet (UV), visible light and infrared (IR). Material maps may be generated using k-means clustering of pixels based on material labels (vegetation, sand, paint, metal, rubber). The clustered object may be positionally matched with the classified object based on locations, orientations and fields of view of the respective cameras. Verification of classified objects may be associated with a certainty score, which may be used downstream for tasks such as trajectory generation.

The application discloses a kind of based on instance image label's correctable package level classification model training method and equipment, the steps of this method are as follows: obtaining panoramic pathology scanning image and corresponding package level label;Image pre-processing and instance image segmentation are carried out to panoramic pathology scanning image;Instance feature package is obtained by using self-supervised learning model to instance image feature extraction;Package level classification model is trained by the package level label and instance feature package obtained;High contribution degree instance image training instance image classifier is screened;The prediction result of package level classification model is corrected using the output result of instance image classifier.The application is suitable for a variety of formats panoramic pathology scanning image classification tasks, can reduce package level classification model repeated training time and instance image level classification error, improve the accuracy and robustness of package level classification model.

The invention discloses a cross-domain hyperspectral image small sample classification method based on a diffusion model, and the method comprises the steps: 1, obtaining a source domain hyperspectral image data set and a target domain hyperspectral image data set, and constructing a source domain training sample set, a target domain training sample set, and a target domain test sample set; step 2, establishing a source domain element task based on the source domain training sample set, and performing joint pre-training on a diffusion-driven domain adapter and an image classifier by using the source domain element task to obtain a source domain pre-training model; step 3, establishing a target domain element task based on the target domain training sample set, and performing fine tuning on the source domain pre-training model by using a target domain training sample to obtain a fine-tuned model; and 4, inputting a target domain test sample into the fine-tuned model, obtaining domain correction features through a domain adapter and a feature extractor, and outputting a classification graph. According to the method, by introducing the diffusion-driven domain adapter, the distribution difference between the source domain and the target domain is relieved, and the cross-domain classification performance under the small sample condition is improved.

The invention relates to the technical field of intelligent algorithm evaluation and optimization and image confrontation attack, in particular to a black box sparse confrontation attack method based on constraint multi-modal multi-target resonance optimization, which comprises the following steps: determining an input image and an image classifier to be confronted, and initializing disturbance optimization parameters; the fitness of constrained multi-modal multi-objective optimization is determined; generating a parent population according to the input image and disturbance optimization parameters; performing clustering, sorting, screening, crossing and variation on a current parent population to generate a child population; when the number of the adversarial samples is greater than a preset number, clustering and redundancy elimination are carried out on disturbance modes of the offspring individuals; sorting union sets of the parent population and the child population, and updating the current parent population; when the maximum query frequency in the disturbance optimization parameters is reached, generating a final confrontation sample based on the disturbance mode of the individuals in the current offspring population; according to the invention, high-quality adversarial samples can be generated.

This invention belongs to the field of deep neural networks and image classification, and specifically relates to an image classification adversarial training improvement method based on boundary sample enhancement. It includes the following steps: Step 1, image sample set preparation; Step 2, designing a GAN network architecture suitable for boundary sample generation; Step 3, training the GAN network suitable for boundary sample generation; Step 4, adversarial training of the image classifier involving boundary samples. This invention effectively constrains the decision boundary stability of the image classification model, enabling the model to perform more refined feature learning in the critical region and avoiding excessive boundary shift due to adversarial perturbations. Simultaneously, it ensures that the model maintains stable and reliable feature representation even when subjected to perturbations. This invention improves model robustness while reducing the damage to the accuracy of natural samples, achieving a better accuracy-robustness balance and stronger generalization ability.