Alerta

IMAGE PROCESSING METHOD AND NEURAL PROCESSING UNIT FOR REGION-SPECIFIC BLURRING

Resumen de: US20260196034A1

0000 An image processing method performed by a neural processing unit is disclosed. The method includes receiving an input image including at least one object and processing the input image using a first model via the neural processing unit to detect a particular object among the at least one object, the first model being an artificial neural network-based object detector trained to detect the particular object in an image. The method further includes processing the input image using a second model via the neural processing unit to blur the particular object in the input image, the second model being an artificial neural network trained to blur a region corresponding to the particular object. An output image including the blurred particular object is generated.

APPARATUS AND NEURAL PROCESSING UNIT FOR EXECUTING A MULTI-INPUT NEURAL NETWORK MODEL

Resumen de: US20260192825A1

A neural processing unit (NPU) includes a controller including a scheduler, the controller configured to receive from a compiler a machine code of an artificial neural network (ANN) including a fusion ANN, the machine code including data locality information of the fusion ANN, and receive heterogeneous sensor data from a plurality of sensors corresponding to the fusion ANN; at least one processing element configured to perform fusion operations of the fusion ANN including a convolution operation and at least one special function operation; a special function unit (SFU) configured to perform a special function operation of the fusion ANN; and an on-chip memory configured to store operation data of the fusion ANN, wherein the schedular is configured to control the at least one processing element and the on-chip memory such that all operations of the fusion ANN are processed in a predetermined sequence according to the data locality information.

SYSTEM AND METHOD PROVIDING A NEURAL NETWORK DRIVEN WIDE NODE SPLIT

Resumen de: US20260195981A1

A computing system and method generate bounding boxes for objects in a three-dimensional scene. A grid generator creates a 3D grid, and a histogram generator builds a histogram of primitive intersections with grid cells. A neural network receives the histogram as input and generates bounding boxes around objects. A box optimizer minimizes the sum of surface areas of the bounding boxes. The histogram can be compressed, using a single bit per bin to indicate primitive presence. Primitives are assigned to child nodes based on proximity or fit. The system provides efficient and optimized bounding box generation for computer graphics applications.

ANOMALY DETECTION USING GRAPH NEURAL NETWORKS

Resumen de: US20260197331A1

0000 Example implementations relate to detecting a terminated entity in a network environment. A network activity dataset including data representative of network activity within a network environment and a plurality of data records is received. Each data record in the plurality of data records includes a set of attributes. A graph that links systems having a first role in the data representative of network activity and a subset of the plurality of data records is generated. Feature information from the set of attributes for one or more data records in the subset of the plurality of data records in the graph is aggregated. A machine learning model is trained based on the aggregated feature information derived from the graph. Using the trained model, a determination representing a likelihood that a respective system having the first role in the data representative of network activity is linked to the terminated entity is generated.

NEURAL NETWORK HIGH RESOLUTION COLOR INFORMATION TO PROJECT IMAGE DATA

Resumen de: US20260196015A1

0000 Apparatuses, systems, and techniques to update lower resolution images. In at least one embodiment, color information from one or more upsampled images may be obtained so that the color information from the one or more upsampled images may be caused to be applied to one or more subsequent lower resolution images.

NAVIGATION METHOD AND APPARATUS, AND VEHICLE

Resumen de: WO2026144120A1

A navigation method and apparatus, and a vehicle. The method (300) comprises: acquiring navigation guidance information, wherein the navigation guidance information indicates at least one road that a vehicle needs to pass through when traveling from the current position to a target position, and a traveling direction associated with each of the at least one road (S310); acquiring road element information, wherein the road element information indicates at least one lane-level element sensed by the vehicle at the current position, the at least one lane-level element indicates a lane boundary and/or a lane centerline of each of M lanes, and the M lanes are lanes in the at least one road (S320); and inputting the navigation guidance information and the road element information into a navigation neural network, so as to obtain navigation field information, wherein the navigation field information indicates passability levels corresponding to different positions of the M lanes (S330). The method can be applied to the field of intelligent driving of intelligent vehicles such as electric vehicles and new energy vehicles, and can reduce, without relying on a high-precision map, the number of ineffective lane changes and/or the probability of deviation during traveling of a vehicle to a destination, thereby improving the traveling efficiency.

BLENDER CONDITIONAL GENERATIVE ADVERSARIAL NETWORKS

Resumen de: US20260196020A1

Certain aspects of the present disclosure provide techniques and apparatus for machine learning. In an example method, a set of exemplars corresponding to a class is accessed, and the set of exemplars is blended to generate a blended exemplar. The blended exemplar is aggregated with a noise sample to generate a noisy exemplar. An output corresponding to the class is generated based on processing the noisy exemplar using a generator neural network. The output is output.

MOVEMENT OF TENSOR DATA DURING RESHAPE OPERATION

Resumen de: US20260195593A1

0000 A method of performing a reshape operation specified in a reshape layer of a neural network model is described. The reshape operation reshapes an input tensor with an input tensor shape to an output tensor with an output tensor shape. The tensor data that has to be reshaped is directly routed between tile memories of the hardware accelerator in an efficient manner. This advantageously optimizes usage of memory space and allows any number and type of neural network models to be run on the hardware accelerator.

MODEL INFERENCE METHOD AND APPARATUS, AND ELECTRONIC DEVICE

Resumen de: WO2026144577A1

Embodiments of the present application provide a model inference method and apparatus, and an electronic device, the method being applicable to a first device, a first model being deployed on the first device, the first model comprising N neural network layers, and the method comprising: determining a type of a first inference task; on the basis of the type of the first inference task, determining, among the N neural network layers, M neural network layers associated with the type of the first inference task, where N and M are each a positive integer greater than or equal to 1, and N is greater than M; and using the M neural network layers to perform inference on first input information, so as to output first output information corresponding to the first input. In the method, the apparatus, or the electronic device, inference is performed on the input information by using only the neural network layers required to be used during inference, and the neural network layers not required to be used are skipped, thus reducing the amount of computation during inference, lowering the power consumption of the device, increasing inference speed, and also saving the operating memory of the device.

METHOD FOR GENERATING FOUNDATION MODEL FOR HYPERSPECTRAL IMAGE BY USING ARTIFICIAL INTELLIGENCE

Resumen de: WO2026146838A1

The present invention relates to a method for generating a foundation model for a hyperspectral image by using artificial intelligence. The method for generating a foundation model for a hyperspectral image by using artificial intelligence: masks a masking patch area randomly determined for a hyperspectral image to be learned; randomly rearranges the positions of some of unmasked patches or adds noise; extracts feature information by inputting the unmasked patches to an encoder; inserts masking feature information into the extracted feature information and inputs same to a decoder to generate a reconstructed hyperspectral image in which the hyperspectral image is reconstructed; and trains the encoder and the decoder so as to minimize the difference between the hyperspectral image and the reconstructed hyperspectral image, thereby making it possible to train and generate a foundation model including an artificial neural network-based encoder and decoder.

DATA SYNCHRONIZATION METHOD AND APPARATUS, AND COMPUTING DEVICE

Resumen de: WO2026145200A1

A data synchronization method and apparatus, and a computing device, which relate to the technical field of computers. The data synchronization method may comprise: for data to be synchronized before execution of a static graph or after execution of the static graph, selecting target data from the data to be synchronized, and performing data synchronization for the target data. The static graph is generated on the basis of a first code segment in a first code, and the first code is code for describing a neural network model. According to the data synchronization method provided by the present application, when data synchronization is performed in the process of executing the code used for describing the neural network model, necessary target data is selected from the data to be synchronized to perform data synchronization, so that the amount of data required to be synchronized in the data synchronization process can be reduced, thereby reducing the consumption of data synchronization, reducing the duration required for data synchronization, and improving the execution efficiency of model training or inference.

RADIOMICS-BASED TREATMENT DECISION SUPPORT FOR LUNG CANCER

Resumen de: US20260192000A1

Two major treatment strategies employed in fighting non-small cell lung cancer (NSCLC) are tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs). The choice of strategy is based on heterogeneous biomarkers expressed by the lung tumor tissue. A major challenge for molecular testing of these biomarkers is the insufficiency of biopsy specimens from patients with advanced NSCLC. Disclosed herein is a method for predicting a response to immune-checkpoint blockade immunotherapy. The method generally involves imaging the subject with positron emission tomography with 2-deoxy-2-fluorine-18 fluoro-D-glucose integrated with computed tomography to produce 18F-FDG PET/CT images of the tumor, analyzing the images using PET, CT, and Kulbek Leibler Divergence statistical (KLD) features or, alternatively using deep leaning such as Neural Networks; generating a radiomic signature from the identified features or Network characteristics; and computing a radiomic score based on the radiomic signature that is predictive of responsiveness to ICIs or TKIs.

Column-Retentive Activation Buffers for Column-Oriented Neural-Network Compute Arrays

Resumen de: US20260195580A1

0000 A column-retentive activation-buffer architecture for column-oriented neural-network compute arrays is disclosed. Each column of processing elements includes a local activation buffer that stores activations received from prior broadcasts for reuse across multiple inference cycles or attention heads, thereby reducing redundant activation transfers from external memory. Lightweight synchronization signals maintain coherence between broadcast control and per-column buffers. The approach preserves CASCADE's row-wise broadcast and column-confined accumulation principles while significantly decreasing external bandwidth and power consumption.

SYSTEM AND METHOD FOR AMBIENT-LIGHT-CORRECTED COMPUTATIONAL PUPILLOMETRY

Resumen de: WO2026147908A1

A computational pupillometry system comprises an imaging device configured to capture video frames of a subject's eye and processors executing instructions to perform advanced pupillary assessment. The system employs multi-frame integration techniques, including super-resolution algorithms that utilize sub-pixel shifts between frames, temporal averaging for noise reduction, and parallax-based artifact mitigation to enhance measurement accuracy. Artificial intelligence models, including temporal neural networks, analyze the enhanced pupillary data to determine pupillary parameters and calculate a light-invariant Pupil Reactivity (PuRe) score. The system processes ambient lighting conditions through computational models that analyze video frames before and after controlled stimulation, enabling consistent scoring across varying environmental conditions. Quality assurance mechanisms provide prerecording and post-recording validation with real-time feedback. The system integrates with electronic medical records through standardized healthcare protocols and supports synchronized, multi-device deployment across healthcare networks.

METHOD FOR GENERATING MULTIMEDIA GAMES AND THEIR LAUNCHING SCHEDULE

Resumen de: WO2026146067A1

The object of the application is a computer-implemented method for generating multimedia games and their launching schedule, in which multimedia game scenarios are made of attributes stored in advance in a database, using a genetic algorithm and deep neural networks, wherein said genetic algorithm is implemented using a bidirectional neural network of the Long Short-Term Memory type, LSTM. The object of the application is also a system for generating multimedia games, said system being configured and programmed to implement the method according to the application.

LOSS SCALING FOR NEURAL NETWORKS

Resumen de: US20260192826A1

A navigation path can be determined for an object using one or more neural networks. In various embodiments, image data is obtained that is representative of an environment in which the object is to be navigated. Relevant features are identified from the image, and a curve fit to those features. Loss values for the potential paths are scaled based at least in part upon the distance of those features in the real world. This can include, in at least some embodiments, performing the scaling as a function of the curvature of the curve fit to the features. Temporal smoothing can be performed with respect to prior path predictions in order to prevent sudden changes in the predicted path. The paths are analyzed to select a path with a highest confidence value that also at least satisfies a minimum confidence criterion. The path can be converted into three-dimensional navigation information.

APPARATUS FOR OBTAINING MAGNETIC RESONANCE IMAGES BASE ON DEEP LEARNING MODEL AND METHOD OF CONTROLLING THE SAME

Resumen de: US20260194613A1

The present disclosure provides an apparatus for restoring the quality of magnetic resonance images based on a deep learning model and a method of controlling the same. The method includes: obtaining a training image corresponding to each magnetic resonance image by applying at least one of a plurality of elements set in connection with the quality of the magnetic resonance image to a magnetic resonance signal corresponding to the magnetic resonance image; obtaining a training dataset including the magnetic resonance image as label data and the obtained training image as input data matching the label data; and training a neural network model based on the training dataset and context data corresponding to the training image. Obtaining the training image includes distorting the magnetic resonance signal by applying the at least one of the plurality of elements and obtaining the training image based on the distorted magnetic resonance signal.

NOISE-RESISTANT COMMUNICATION METHOD AND APPARATUS BASED ON FULLY CONNECTED NEURAL NETWORK

Resumen de: US20260195566A1

The present application relates to a noise-resistant communication method and apparatus based on a fully connected neural network. The method includes: encoding, before signal transmission, target information based on encoding sequences to obtain and transmit a target signal, and receiving, upon signal transmission, the target signal, and decoding the target signal via a fully connected neural network to obtain the target information, where the fully connected network takes the target signal as input and the target information as output, and a hidden layer of the fully connected neural network includes a plurality of fully connected layers for decoding. The method decodes a transmitted signal affected by noise interference via the neural network, so that a problem of an increased bit error rate caused by strong noise interference in the transmitted signal is effectively avoided.

METHOD FOR PROCESSING AND ANALYZING DATA INCLUDING HYPERSPECTRAL IMAGE ON BASIS OF ARTIFICIAL GENERAL INTELLIGENCE (AGI) AND FOUNDATION MODEL (FM)

Resumen de: WO2026146839A1

The present invention relates to a method for processing and analyzing data including a hyperspectral image on the basis of artificial general intelligence (AGI) and a foundation model (FM) and, more specifically, to a method for processing and analyzing data including a hyperspectral image on the basis of artificial general intelligence (AGI) and a foundation model (FM), the method comprising: receiving a request for a task to be performed on a hyperspectral image from a user; determining, using an LLM, whether the task needs to be processed by a first task processing module based on deep learning or a second task processing module based on a rule base; and when the hyperspectral image needs to be processed by the first task processing module based on deep learning, inputting feature information extracted by inputting the hyperspectral image to an encoder into a first sub-task module based on an artificial neural network to derive a task result for the task requested by the user to be performed on the hyperspectral image.

SYSTEM, METHOD AND COMPUTER-ACCESSIBLE MEDIUM FOR FACILITATING CANCER SCREENING AND RISK PROFILING

Resumen de: US20260196353A1

0000 Exemplary systems, methods and Computer-accessible medium according to the exemplary embodiments of the present disclosure can provide a Multi-modal Transformer (MMT), a neural network that synergistically utilizes mammography and ultrasound to identify existing cancers and estimate future cancer risk. MMT aggregates multi-modal data through self-attention and modeling temporal tissue changes by comparing current exams to prior imaging. Thus, exemplary method, system and computer-accessible medium can be provided for detecting cancer. with which it possible to receive, with an artificial intelligence (AI) procedure, a plurality of scanning images associated with multiple modalities for at least one portion of a body, train the AI procedure on a multi-modal image dataset based on the plurality of scanning images, and predict, by the trained AI procedure, an existence of the cancer based on the multiple modalities of the plurality of scanning images.

METHOD AND DEVICE FOR ENABLING AN EXPLANATION AND/OR VERIFICATION OF A BEHAVIOUR OF A NEURAL NETWORK TRAINED WITH A TRAINING DATABASE, AND VEHICLE WITH AI

Resumen de: WO2025045319A1

The invention relates to a method for explaining and/or verifying a behaviour of a neural network trained with a training database, said method having the following steps: - determining sensor information by means of a sensor, so that sensor information is available, - applying the neural network to the sensor information, so that target information is available, - determining similarity information, wherein a comparison is made of the sensor information or part of the sensor information with an information database, so that the similarity information is available, linking the similarity information with the target information, so that a target information tuple is available, on the basis of which the explanation and/or verification of the behaviour of the neural network can be realised.

SYSTEM AND METHOD FOR PREDICTING FORMATION IN SPORTS

Resumen de: EP4773045A2

0001 A system and method of predicting a team's formation on a playing surface are disclosed herein. A computing system retrieves one or more sets of event data for a plurality of events. Each set of event data corresponds to a segment of the event. A deep neural network, such as a mixture density network, learns to predict an optimal permutation of players in each segment of the event based on the one or more sets of event data. The deep neural network learns a distribution of players for each segment based on the corresponding event data and optimal permutation of players. The computing system generates a fully trained prediction model based on the learning. The computing system receives target event data corresponding to a target event. The computing system generates, via the trained prediction model, an expected position of each player based on the target event data.

Validation device of quantum self-attention neural networks for image classification and method for controlling the same

Nº publicación: KR20260106911A 07/07/2026

Solicitante:

연세대학교산학협력단

Resumen de: KR20260106911A

본 개시는 이미지 분류를 위한 양자 셀프 어텐션 뉴럴 네트워크 검증 장치 및 그 제어 방법에 관한 것으로, 상기 입력 모듈을 통하여 이미지를 수집하고, 수집된 이미지를 소정의 패치 단위로 분리하고, 분리된 상기 이미지 패치를 인코딩하고, 인코딩된 상기 이미지 패치를 기초로 쿼리, 키 및 값을 계산하고, 상기 이미지 패치를 벡터로 변환시키는 방법을 사용하여 개별 이미지 패치의 어텐션 계수를 계산하고, 상기 이미지 패치의 상기 어텐션 계수의 가중치 합을 이용하여 출력값을 계산하고, 시컨스에 걸쳐 상기 이미지 패치의 어텐션 계수의 평균을 구하고, 상기 이미지 패치의 어텐션 맵의 규준화를 수행하고, 상기 이미지 패치의 상기 쿼리 키 어텐션 계수의 행별(row-wise) 합을 계산하고, 상기 이미지 패치의 상기 행별 합 결과를 이미지 패치 형태로 변환하고, 변환된 상기 이미지 패치를 원본 이미지에 맞게 조정하는 것을 그 요지로 한다.