Alerta

PERFORMING SEMANTIC SEGMENTATION OF 3D DATA USING DEEP LEARNING

Resumen de: US20260311336A1

0000 A computer-implemented method of training a deep artificial neural network includes receiving a three-dimensional point cloud and training the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network. A computing system includes a processor; and a memory having stored thereon computer-executable instructions that, when executed by the processor, cause the computing system to receive a three-dimensional point cloud and train the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network. In yet another aspect, a non-transitory computer-readable medium includes computer-executable instructions that when executed, cause a computer to receive a three-dimensional point cloud and train the deep artificial neural network by subdividing the three-dimensional point cloud, and updating weights of the deep artificial neural network.

SYSTEMS AND METHODS FOR ROPE EVALUATION

Resumen de: US20260309639A1

0000 In an embodiment of the disclosed principles, a method of analyzing a rope to estimate its residual break strength (RBS) is provided. The method entails training a multi-layered neural network to recognize rope damage and estimate an RBS for a rope under study by collecting high-resolution visual data of a plurality of sample ropes, extracting visual features of damaged areas of each rope, resolving each visual feature into a damage type and clustering damage type exemplars, breaking each sample rope to determine an actual RBS for each rope and classifying the visual data via determined RBS, specific to product type/damage mode. Job and product data are entered for the rope under study, the rope is paid out, and high-resolution visual data is captured while spooling back the rope. The correlated data is provided to the multi-layered neural network to generate an estimate of RBS for the rope under study.

OPTICAL INFORMATION READING DEVICE

Resumen de: US20260310918A1

To suppress an increase in processing time due to a load of inference processing while improving reading accuracy by the inference processing of machine learning. An optical information reading device includes a processor including: an inference processing part that inputs a code image to a neural network and executes inference processing of generating an ideal image corresponding to the code image; and a decoding processing part that executes first decoding processing of decoding the code image and second decoding processing of decoding the ideal image generated by the inference processing part. The processor executes the inference processing and the first decoding processing in parallel, and executes the second decoding processing after completion of the inference processing.

OBJECT IMAGE COMPLETION

Resumen de: US20260311379A1

Apparatuses, systems, and techniques to generate complete depictions of objects based on a partial depiction of the object. In at least one embodiment, an image of a complete object is generated by one or more neural networks, based on an image of a portion of the object, using an encoder of the one or more neural networks trained using training data generated from output of a decoder of the one or more neural networks.

VIEW GENERATION USING ONE OR MORE NEURAL NETWORKS

Resumen de: US20260311378A1

Apparatuses, systems, and techniques are presented to generate image or video content representing at least one point of view. In at least one embodiment, one or more neural networks are used to generate one or more images of one or more objects from a first point of view based at least in part upon one or more images of the one or more objects from a second point of view.

Neural Network Initialization

Resumen de: US20260310969A1

0000 A neural network representing a controlled space can be initialized by collecting state time series data that affects the controlled space such as weather, and also collecting sensor data from the controlled space at the same time. The time series data is used as input to a neural network that models the controlled space until an area in the neural network equivalent to the sensor is at or near the sensor state at a given time.

PANORAMA GENERATION USING ONE OR MORE NEURAL NETWORKS

Resumen de: US20260311290A1

0000 Apparatuses, systems, and techniques are presented to generate panoramas from individual images. In at least one embodiment, one or more generative neural networks are used to generate a spherical panoramic image using features extracted from a single input image.

Identification of multi-scale features using neural network

Resumen de: AU2026234120A1

Apparatuses, systems, and techniques to identify features within one or more images. In at least one embodiment, features are identified in one or more images using one or more neural networks containing convolutional layers with multiple filters that may be executed by one or more parallel processing unit. ep e p

TRAINING VISION LANGUAGE NEURAL NETWORKS WITH COMPONENT REUSE

Resumen de: US20260311476A1

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating and training a glued neural network that reuses components taken from a first pre-trained neural network and a second pre-trained neural network. For example, the first pre-trained neural network can be a multi-modal neural network that includes a visual encoder neural network and a small-scale language model neural network. The second pre-trained neural network can be a large-scale language model neural network.

TECHNIQUES FOR ARTIFICIAL INTELLIGENCE CAPABILITIES AT A NETWORK SWITCH

Resumen de: US20260312885A1

0000 Examples include techniques for artificial intelligence (AI) capabilities at a network switch. These examples include receiving a request to register a neural network for loading to an inference resource located at the network switch and loading the neural network based on information included in the request to support an AI service to be provided by users requesting the AI service.

DEPTH AND MOTION ESTIMATIONS IN MACHINE LEARNING ENVIRONMENTS

Resumen de: US20260311478A1

A mechanism is described for facilitating depth and motion estimation in machine learning environments, according to one embodiment. A method of embodiments, as described herein, includes receiving a frame associated with a scene captured by one or more cameras of a computing device; processing the frame using a deep recurrent neural network architecture, wherein processing includes simultaneously predicating values associated with multiple loss functions corresponding to the frame; and estimating depth and motion based the predicted values.

Systems and Methods for Direct-From-Non-Volatile-Memory Neural Network Inference and Hidden-State Reuse

Resumen de: US20260310378A1

0000 Systems and methods for performing inference with a neural network model using model weights stored in non-volatile memory without loading the model weights into dynamic random-access memory (DRAM) prior to inference. A processing system reads model weights from non-volatile memory, streams the model weights to a compute engine, generates outputs by performing multiply-accumulate operations using the streamed model weights, and maintains intermediate states in on-chip volatile memory. In some embodiments, a recurrent computational model maintains a hidden state with fixed dimensionality independent of sequence length. In some embodiments, the processing system pre-processes context data to generate hidden states, stores the hidden states, and retrieves the hidden states to resume inference without reprocessing the context data. The hidden states may be stored as tuples comprising a key and a hidden state in a vector database.

DYNAMIC PARALLELISM

Resumen de: US20260311000A1

0000 In various examples, one or more neural networks are to be performed by a number of processors. The neural network may be reconfigured during inferencing such that the number of processors that start the inferencing is different from the number of processors that finish the inferencing.

TEMPORAL FEATURE SCALING FOR NEURAL NETWORK PROCESSING

Resumen de: US20260311361A1

Systems, devices, and methods described herein relate to feature scaling for neural network processing. In some examples, temporal feature scaling is employed for efficient neural network processing. To process the neural network, a processing system accesses feature maps of a current input frame in a temporal sequence of input frames and scales the feature maps of the current input frame using excitation values generated from a previous input frame in the temporal sequence. The processing system generates new excitation values from the current input frame to perform scaling for a subsequent input frame in the temporal sequence.

High Likeness High Compliance Human Figure Generation

Resumen de: US20260311382A1

Techniques relating to high likeness high compliance human figure generation are disclosed. A method for high likeness high compliance human figure generation may include selecting training images as a basis for an expert learning model, obtaining embeddings for the training images, the embedding being recorded separately for each expert, computing a geometric property of the embedding space for each expert, embedding a received prompt using the same embedding method as for the training images, and generating an output image using a subset of experts selected for highest compliance (e.g., most similarity) to the received prompt, or for other purposes. A system for implementing this method may include a training data ensemble, an image augmentation pipeline, an LLM and other neural networks, a mood and style ensemble of lightweight expert models built on a foundational model, and a tokenizer.

MISSION PREDICTION SYSTEM FOR SIMULTANEOUS MISSION SUCCESS AND ENEMY INTENT PREDICTION USING GRAPH NEURAL NETWORKS

Resumen de: US20260310955A1

The mission prediction system models a wargaming scenario as a graph. Once the GNN graph is defined, multiple sets of AIGC data is used to train the GNN graph. The trained GNN graph can then be used to perform inference on novel wargaming scenarios to simultaneously predict mission success and enemy intent in real-time, even if the novel wargaming scenarios involve a much greater number of assets than were used in the AIGC data. A viewer can be used to review the predictions and to analyze the novel wargaming scenario at different timesteps.

Corrosion Type Identification in Flanged Joints Using a Novel Recurrent Neural Network Technique Applied to Electrochemical Noise Measurements

Resumen de: US20260309616A1

A method is provided for automatically determining corrosion-related temporal features using an autoencoder that considers and has as an input electrochemical noise (EN) data from a metallic structure, wherein the determined corrosion-related temporal features are used to classify and/or characterize types of corrosion of the metallic structure. The method relies on the autoencoder and EN data to automatically determine the corrosion-related temporal features without requiring prior labeling of data or human supervision. The disclosed method may be implemented in a computer program product or an apparatus that enable classification or characterization of types of corrosion of the metallic structure. In some embodiments, the method may include the step of measuring the raw EN data and the apparatus may include an EN measurement device for measuring raw EN data.

DEEPFAKE SHIELDING OF IMAGES

Resumen de: WO2026213578A1

Embodiments of the present disclosure include techniques for processing images that depict human faces. In one embodiment, one or more source images comprising a source identity are received. At runtime, a source image is applied to a neural network to generate adversarial pixel signals to be applied to the source image. The source image to generate a shielded image that retains the source identity but with modified pixels depicting the face of the person. The shielded image is provided by the image processing system. Applying the shielded image to a face swapper produces a swap identity that is visually distinct from the source identity. The visual distinctions between the swap identity and the source identity are caused by an adversarial attack that emerges in a facial detection component, a landmark detection component, and/or an identity mapping component of the face swapper.

Video Compression Method and Device with Multi-Scale Implicit Neural Representation

Resumen de: US20260311377A1

0000 A video processing method with multi-scale implicit neural representation includes: using an implicit neural network to convert an original video sequence into network parameters; using the implicit neural network to extract fused encoding features with temporal information and perform spatial decoding, where the fused encoding features are selected based on the original video sequence; constructing a loss function of the implicit neural network based on the original video sequence and a reconstructed video, where the reconstructed video is obtained by reconstructing the original video sequence; using the loss function to iteratively optimize the network parameters, where the final optimized network parameters are used as the video compression results corresponding to the original video sequence. Therefore, by adopting the video processing method, the spatial and temporal structural priors of the video can be fully leveraged, enabling the compression efficiency to reach optimal levels and improving video compression performance.

ASSEMBLY LINE VARIANT PATH PLANNING SYSTEM AND METHOD BASED ON GRAPH-BASED DEEP REINFORCEMENT LEARNING

Resumen de: US20260311074A1

An assembly line variant path planning system based on graph-based deep reinforcement learning includes an input layer, a knowledge and mechanism layer, a core processing module, an output layer, and an application and feedback layer. An assembly line variant process is modelled as a graph structure and a reward function that precisely matches a change propagation preference objective is designed. By using the input change propagation preference mathematical model as a training basis, a neural network model constructed based on graph-based deep reinforcement learning is repeatedly trained until a model output result is stabilized and converged, thereby intelligently identifying an optimal preliminary variant path. A key dimensional linkage skeleton for subsequent instantiated reconstruction is provided by this optimal path, ultimately enabling rapid virtual-physical mapping modeling of the entire assembly process and significantly reducing time for physical assembly line reverse modeling.

OPTICAL SENSING WITH NONLINEAR OPTICAL NEURAL NETWORKS

Resumen de: US20260310953A1

0000 Methods, devices, and systems for optical sensing with nonlinear optical neural network (ONNs) are provided. In one aspect, a method includes: receiving light from a visual scene by a first optical linear layer in a nonlinear ONN apparatus, linearly transforming the light into first optical outputs by the first optical linear layer trained to perform a first optical linear operation, non-linearly generating second optical outputs based on the first optical outputs by an optical nonlinear layer in the ONN apparatus, and linearly transforming the second optical outputs into third optical outputs by a second optical linear layer in the ONN apparatus trained to perform a second optical linear operation. The first optical linear layer, the optical nonlinear layer, and the second optical linear layer are sequentially arranged in series in the ONN apparatus.

CONVOLUTIONAL NEURAL NETWORK STRUCTURE FOR EXPLAINABLE 3D SHAPE LEARNING, AND METHOD AND SYSTEM FOR 3D SHAPE LEARNING USING SAME

Resumen de: US20260311451A1

0000 Disclosed are a convolutional neural network structure for explainable 3D shape learning, and a method and a system for 3D shape learning by using same. A method for shape learning performed by a shape learning system according to an embodiment may comprise: extracting geodesic features and geometric features from three-dimensional shape data; and classifying an object through a convolution operation on the extracted geodesic features and the extracted geometric features with respect to faces constituting the three-dimensional shape data.

MODULARIZED HOME ROBOT

Resumen de: US20260307239A1

0000 Implementations of a fully configurable, modularized home robot are described that implement trained neural network classifiers to solve problems of providing home health care and monitoring of the elderly and/or infirm, providing entertainment for the family, providing environmental and safety monitoring coupled with mechanisms to clean the air and remedy indoor climates such as humidity and temperature, and provide a mechanism for caring for pets left at home when the owner is away based upon sensory input.

SYSTEMS AND METHODS FOR IMPROVED IMAGE PROCESSING

Resumen de: US20260311475A1

0000 A system and method for training (and inference) of a NN model for wide/ultra-wide angle lens object detection and positioning, the NN training module configured to receive one or more images associated with one or more images from the object recognition data set, the data set including one or more bounding box labels associated with the one or more images, apply a customized data augmentation routine on said images to create one or more augmented images in each new iteration of training, provide a predetermined neural network for computer vision model to test and train the neural network for computer vision, predict and obtain a bounding box and confidence score, each being associated with one or more objects in the one or more images from the neural network for computer vision, match ground truth bounding boxes in image space to augmentations from a box data store, apply regression loss on the bounding boxes and compute the cost between predicted values and the label values to determine change in weightings.

VESTIBULAR FUNCTION TESTING SYSTEM, DATA PROCESSING METHOD, ELECTRONIC DEVICE, AND PROGRAM PRODUCT

Nº publicación: WO2026208938A1 08/10/2026

Solicitante:

SHANGHAI ZEHNIT MEDICAL TECH CO LTD [CN]
\u4E0A\u6D77\u5FD7\u542C\u533B\u7597\u79D1\u6280\u6709\u9650\u516C\u53F8

Resumen de: WO2026208938A1

A vestibular function testing system, a data processing method, an electronic device, and a program product. The method comprises: acquiring an eye movement video of a subject during vestibular function testing; extracting an eye movement trajectory of the subject from the eye movement video, the eye movement trajectory comprising at least one of the following: an eye movement trajectory in a horizontal direction, an eye movement trajectory in a vertical direction, and an eye movement trajectory in a torsional direction; acquiring auxiliary test data of the subject, the auxiliary test data comprising body posture data of the subject during the vestibular function testing and/or nystagmus characteristic parameters acquired from the eye movement trajectory; and inputting the eye movement trajectory and the auxiliary test data into a trained neural network model, and identifying and outputting vestibular function data of the subject by means of the neural network model, wherein the eye movement trajectory and the auxiliary test data are subjected to time synchronization and association. Compared with a solution in which only an eye movement video or eye movement trajectory data is considered, the method further adds the body posture data and/or the nystagmus characteristic parameters, so that an evaluation result of vestibular function is more accurate.