Contextual embeddings for improving static analyzer output

In a computer-implemented method for improving a static analyzer output, a processor receives a labeled data set with labeled true vulnerabilities and labeled false vulnerabilities. A processor receives pretrained contextual embeddings from a contextual embeddings model. A processor maps the true vulnerabilities and the false vulnerabilities to the pretrained contextual embeddings model. A processor generates a fine-tuned model with classifications for true vulnerabilities.

Convolutional computing using multilayered analysis engine

Disclosed embodiments provide for deep convolutional neural network computing. The convolutional computing is accomplished using a multilayered analysis engine. The multilayered analysis engine includes a deep learning network using a convolutional neural network (CNN). The multilayered analysis engine is used to analyze multiple images in a supervised or unsupervised learning process. Multiple images are provided to the multilayered analysis engine, and the multilayered analysis engine is trained with those images. A subject image is then evaluated by the multilayered analysis engine. The evaluation is accomplished by analyzing pixels within the subject image to identify a facial portion and identifying a facial expression based on the facial portion. The results of the evaluation are output. The multilayered analysis engine is retrained using a second plurality of images.

Using iterative 3D-model fitting for domain adaptation of a hand-pose-estimation neural network

Described is a solution for an unlabeled target domain dataset challenge using a domain adaptation technique to train a neural network using an iterative 3D model fitting algorithm to generate refined target domain labels. The neural network supports the convergence of the 3D model fitting algorithm and the 3D model fitting algorithm provides refined labels that are used for training of the neural network. During real-time inference, only the trained neural network is required. A convolutional neural network (CNN) is trained using labeled synthetic frames (source domain) with unlabeled real depth frames (target domain). The CNN initializes an offline iterative 3D model fitting algorithm capable of accurately labeling the hand pose in real depth frames. The labeled real depth frames are used to continue training the CNN thereby improving accuracy beyond that achievable by using only unlabeled real depth frames for domain adaptation.

Master transform architecture for deep learning

Apparatuses, systems, and techniques to transform input data for training neural networks. In at least one embodiment, one or more data transforms are identified in a sequence of data transforms and combined into one or more master data transforms to be performed by one or more parallel processing units in order to prepare data for training an untrained neural network.

Machine-learning based gesture recognition using multiple sensors

A device implementing a system for machine-learning based gesture recognition includes at least one processor configured to, receive, from a first sensor of the device, first sensor output of a first type, and receive, from a second sensor of the device, second sensor output of a second type that differs from the first type. The at least one processor is further configured to provide the first sensor output and the second sensor output as inputs to a machine learning model, the machine learning model having been trained to output a predicted gesture based on sensor output of the first type and sensor output of the second type. The at least one processor is further configured to determine the predicted gesture based on an output from the machine learning model, and to perform, in response to determining the predicted gesture, a predetermined action on the device.

DATA GATHERING AND DATA SELECTION TO TRAIN A MACHINE LEARNING ALGORITHM

Disclosed are techniques for training a position estimation module. In an aspect, a first network entity obtains a plurality of positioning measurements, obtains a plurality of positions of one or more user equipments (UEs), the plurality of positions determined based on the plurality of positioning measurements, stores the plurality of positioning measurements as a plurality of features and the plurality of positions as a plurality of labels corresponding to the plurality of features, and trains the position estimation module with the plurality of features and the plurality of labels to determine a position of a UE from positioning measurements taken by the UE.

System and method for data augmentation for trace dataset

A system and method for classification. In some embodiments, the method includes forming a first training dataset and a second training dataset from a labeled input dataset; training a first classifier with the first training dataset; training a variational auto encoder with the second training dataset, the variational auto encoder comprising an encoder and a decoder; generating a third dataset, by feeding pseudorandom vectors into the decoder; labeling the third dataset, using the first classifier, to form a third training dataset; forming a fourth training dataset based on the third dataset; and training a second classifier with the fourth training dataset.

Multi-agent coordination method and apparatus

The present disclosure discloses a multi-agent coordination method. The method includes: performing multiple data collections on N agents to collect E sets of data, where N and E are integers greater than 1; and optimizing neural networks of the N agents using reinforcement learning based on the E sets of data. Each data collection includes: randomly selecting a first coordination pattern from multiple predetermined coordination patterns; obtaining N observations after the N agents act on an environment in the first coordination pattern; determining a first probability and a second probability that a current coordination pattern is the first coordination pattern based on the N observations; and determining a pseudo reward based on the first probability and the second probability. The E sets of data include: a first coordination pattern label indicating the first coordination pattern, the N observations, and the pseudo reward.

TRAILER HITCHING ASSIST SYSTEM WITH TRAILER COUPLER DETECTION

A vehicular trailer hitching assist system includes a camera disposed at a rear portion of a vehicle and viewing at least rearward of the vehicle. During a reversing maneuver of the vehicle toward a trailer that is spaced from the vehicle at a distance from the vehicle, the camera views at least a portion of a front profile of the trailer. An electronic control unit (ECU) includes an image processor operable to process image data captured by the camera. The vehicular trailer hitching assist system, via image processing at the ECU of image data captured by the camera during the reversing maneuver of the vehicle toward the trailer, determines a plurality of landmarks corresponding to the front profile of the trailer. Based at least in part on the determined plurality of landmarks, the vehicular trailer hitching assist system determines location of a trailer coupler of the trailer.

Methods and systems for mining minority-class data samples for training a neural network

Methods and systems for mining minority-class data samples are described. A minority-class mining service receives activations generated by an inner-layer of a client neural network that has been trained to perform a prediction task that involves classification. The minority-class mining service generates a recalibrated activation using a recalibration neural network, and generates an anomaly detector output using an anomaly detector. From the anomaly detector output, a minority-class score is computed for the data sample represented by a received activation. The computed minority-class score is compared against a minority-class threshold to identify a candidate minority-class data sample. The candidate minority-class data sample can then be labeled and added to the training dataset for the client neural network.

Anomaly detection in event-based systems using image processing

At least one processor may capture a plurality of image snapshots containing information about a monitored system at a plurality of sequential times, each snapshot having the same vertical and horizontal dimensions. The processor may label the plurality of image snapshots as indicative of an event that took place in the monitored system, may receive additional data describing the event, may cluster the labeled plurality of image snapshots and the additional data using at least one machine learning clustering algorithm, and may merge the clustered plurality of image snapshots and the clustered additional data into merged data. The processors may create a model by processing the merged data using at least one neural network, the model being configured to detect future events of a same type as the event in the monitored system. The processor may store the model in a memory in communication with the processor.

Adversarial bootstrapping for multi-turn dialogue model training

Systems described herein may use machine classifiers to perform a variety of natural language understanding tasks including, but not limited to multi-turn dialogue generation. Machine classifiers in accordance with aspects of the disclosure may model multi-turn dialogue as a one-to-many prediction task. The machine classifier may be trained using adversarial bootstrapping between a generator and a discriminator with multi-turn capabilities. The machine classifiers may be trained in both auto-regressive and traditional teacher-forcing modes, with the maximum likelihood loss of the auto-regressive outputs being weighted by the score from a metric-based discriminator model. The discriminators input may include a mixture of ground truth labels, the teacher-forcing outputs of the generator, and/or negative examples from the dataset. This mixture of input may allow for richer feedback on the autoregressive outputs of the generator. Additionally, dual sampling may improve response relevance and ...

Regression-based line detection for autonomous driving machines

In various examples, systems and methods are disclosed that preserve rich spatial information from an input resolution of a machine learning model to regress on lines in an input image. The machine learning model may be trained to predict, in deployment, distances for each pixel of the input image at an input resolution to a line pixel determined to correspond to a line in the input image. The machine learning model may further be trained to predict angles and label classes of the line. An embedding algorithm may be used to train the machine learning model to predict clusters of line pixels that each correspond to a respective line in the input image. In deployment, the predictions of the machine learning model may be used as an aid for understanding the surrounding environment—e.g., for updating a world model—in a variety of autonomous machine applications.

Unsupervised contrastive learning of visual representations using negative mixing

A training system includes: an encoder module configured to receive a query image and to generate a first vector representative of one or more features in the query image using an encoder; a mixing module configured to generate a second vector by mixing a third vector, representative of one or more features in a second image that is classified as a negative relative to the query image, with a fourth vector; and an adjustment module configured to train the encoder by selectively adjusting one or more parameters of the encoder based on the first vector and the second vector.

Optimized policy-based active learning for content detection

Systems and methods for training an object detection network are described. Embodiments train an object detection network using a labeled training set, wherein each element of the labeled training set includes an image and ground truth labels for object instances in the image, predict annotation data for a candidate set of unlabeled data using the object detection network, select a sample image from the candidate set using a policy network, generate a labeled sample based on the selected sample image and the annotation data, wherein the labeled sample includes labels for a plurality of object instances in the sample image, and perform additional training on the object detection network based at least in part on the labeled sample.

Automating complex processes

Automating a process to process a task or an object involves defining elements of the process in one or more human-intelligible and editable and machine interpretable workflow program documents. The documents each include a plurality of actors who perform actions or take decisions, a sequence of action steps each associated with an actor and having at least one expected outcome and a corresponding next step for each expected outcome. The process is executed by a processor running the code defined by the documents. If an exception is detected in the processing of a task or object according to the code, the exception is passed to a supervisory function to perform a remedial action. Remedial actions include individually modifying the task or object to be processed or on the fly patching or modifying of the documents. The modified documents are interpreted or re-compiled and the processor subsequently executes a modified process according to the modified document.

Feature value generation device, feature value generation method, and program

A feature value generation device includes a generator configured to digitize non-numerical text data items collected at a plurality of timings from a target of anomaly detection, to generate vectors whose elements are feature values corresponding to the digitized data items; a learning unit configured to learn the vectors during a learning period so as to output a learning result; and a detector configured to detect, during a test period, for each of the vectors generated by the generator, an anomaly based on said each of the vectors and the learning result.

System and method for machine-learning based extraction of information from documents

A method and system are provided for training a machine-learning (ML) system/module and to provide an ML model. In one embodiment, a method includes using a labeled entities set to train a machine learning (ML) system, to obtain an ML model, and using the trained ML model to predict labels for entities in an unlabeled entities set, yielding a machine-labeled entities set. One or more individual ML models may be trained and used in this way, where each individual ML model corresponds to a respective document source. The document sources can be identified via classification of a corpus of documents. The prediction of labels provides a respective confidence score for each machine-labeled entity. The method also includes selecting from the machine-labeled entities set, a subset of machine-labeled entities having a respective confidence score at least equal to a threshold confidence score; and updating the labeled entities set by adding thereto the selected subset of machine-labeled entities ...

Learning contrastive representation for semantic correspondence

A multi-level contrastive training strategy for training a neural network relies on image pairs (no other labels) to learn semantic correspondences at the image level and region or pixel level. The neural network is trained using contrasting image pairs including different objects and corresponding image pairs including different views of the same object. Conceptually, contrastive training pulls corresponding image pairs closer and pushes contrasting image pairs apart. An image-level contrastive loss is computed from the outputs (predictions) of the neural network and used to update parameters (weights) of the neural network via backpropagation. The neural network is also trained via pixel-level contrastive learning using only image pairs. Pixel-level contrastive learning receives an image pair, where each image includes an object in a particular category.

Methods and apparatus for generating training data to train machine learning based models

Training data is obtained, wherein the training data comprises labelled samples and unlabeled samples 802. Clusters of the training data are generated based on one or more corresponding attributes of the training data 804. A distance metric between positively labelled samples and unlabelled samples within each cluster is determined 806, and for each of the clusters, a plurality of sub-clusters based on the determined distance metrics is determined 808. One or more of the unlabelled samples from each of the plurality of sub-clusters is determined based on a corresponding reward value and a corresponding sampling rate value 810. The determined unlabelled samples are then stored 812. The stored samples may then be labelled and used to improve the training of a machine learning model.

Systems and methods for fine tuning image classification neural networks

An authentication engine, residing at one or more computing machines, receives, from a vision device comprising one or more cameras, a probe image. The authentication engine generates, using a trained facial classification neural engine, one or more first labels for a person depicted in the probe image and a probability for at least one of the one or more first labels. The authentication engine determines that the probability is within a predefined low accuracy range. The authentication engine generates, using a supporting engine, a second label for the person depicted in the probe image. The supporting engine operates independently of the trained facial classification neural engine. The authentication engine further trains the facial classification neural engine based on the second label.

MACHINE LEARNING METHODS AND SYSTEMS FOR VARIETY PROFILE INDEX CROP CHARACTERIZATION

A system includes one or more processors; and one or more non-transitory, computer-readable media including instructions that, when executed by the one or more processors, cause the computing system to: receive a machine data set; process the machine data set with a trained machine-learned model to generate predicted variety profile index values; and cause a visualization to be displayed. A computer-implemented method includes receiving a machine data set; processing the machine data set with a trained machine-learned model to generate predicted variety profile index values; and causing a visualization to be displayed. A non-transitory computer-readable medium includes computer-executable instructions that, when executed by one or more processors, cause a computer to: receive a machine data set; process the machine data set with a trained machine-learned model to generate predicted variety profile index values; and cause a visualization to be displayed.

Systems and methods for contrastive learning of visual representations

Systems, methods, and computer program products for performing semi-supervised contrastive learning of visual representations are provided. For example, the present disclosure provides systems and methods that leverage particular data augmentation schemes and a learnable nonlinear transformation between the representation and the contrastive loss to provide improved visual representations. Further, the present disclosure also provides improvements for semi-supervised contrastive learning. For example, computer-implemented method may include performing semi-supervised contrastive learning based on a set of one or more unlabeled training data, generating an image classification model based on a portion of a plurality of layers in a projection head neural network used in performing the contrastive learning, performing fine-tuning of the image classification model based on a set of one or more labeled training data, and after performing the fine-tuning, distilling the image classification model ...

Methods for using machine learning and mechanistic models for biological feature mapping with multiparametric MRI

Described here are systems and methods for generating and implementing a hybrid machine learning and mechanistic model to produce biological feature maps, or other measurements of biological features, based on an input of multiparametric magnetic resonance or other images. The hybrid model can include a combination of a machine learning model and a mechanistic model that takes as an input multiparametric MRI, or other imaging, data to generate biological feature maps (e.g., tumor cell density maps), or other measures or predictions of biological features (e.g., tumor cell density). The hybrid models have capabilities of learning individual-specific relationships between imaging features and biological features.

System and method for validating data

A system and method are provided for validating data. The method is executed by a device having a data interface coupled to a processor and includes obtaining a validation set comprising at least one validation case, each validation case comprising at least one test condition. The method also includes obtaining, via the data interface, at least one data set to be validated using the validation set. The method also includes applying the validation set to the at least one data set to validate the data in the data set by, for each record in the at least one data set, validating a value in the record according to the at least one test condition. The method also includes outputting a validation result for each record.

Automating complex document handling processes

Software for automating a document handling process, such as sending a contract for signatures. The document handling process is defined in a tabular workflow program document, such as a spreadsheet. Each row represents a path through a portion of the document handling process to an outcome. Each column holds a human readable label for a process component, such as a process portion name, an actor for carrying out that portion, an action to be carried out, an outcome, and a next step. The software processes or filters the workflow program document to visualise and implement the workflow using an existing document handling visual programming framework. The workflow is visualised as a flowchart using pre-built visual representations of workflow components. The workflow is implemented by programming associated with these visual representations. In this way, the workflow program document directs a portion of the document handling process along one of the paths to an outcome based on the human ...

Active learning for inspection tool

METHOD AND SYSTEM FOR CONTEXT-AWARE DECISION MAKING OF AN AUTONOMOUS AGENT

A system for context-aware decision making of an autonomous agent includes a computing system having a context selector and a map. A method for context-aware decision making of an autonomous agent includes receiving a set of inputs, determining a context associated with an autonomous agent based on the set of inputs, and optionally any or all of: labeling a map; selecting a learning module (context-specific learning module) based on the context; defining an action space based on the learning module; selecting an action from the action space; planning a trajectory based on the action S260; and/or any other suitable processes.

Identifying patterns by applying a machine learning algorithm to the automation of complex processes

Software for identifying patterns in an automated process, such as a document handling process, manufacturing process, creative project, or software development project. The execution history of processing a workflow program document is recorded in a log document. A machine learning algorithm processes the log document to identify a pattern. The identified pattern is used to make a workflow suggestion. The suggestion may comprise: a sequence of steps to perform a further component of an existing process; a sequence of steps to perform a component of a further process; identification of similar sequences of steps to validate differences between the sequences; a suggested modification of similar sequences for consistency; identification of a possible error in a sequence of steps; a suggested modification to mitigate an identified error or exception; or a suggested template for creating future workflow program documents. The workflow program document may be a spreadsheet with rows representing ...

Encoding processes to execute complex workflows

Encoding a program defining a workflow for implementing a process involves defining a sequence of actions by a series of first rows in a table. For each row, the following are identified: 1. an actor from a set of actors associated with an action in a first column; 2. an action step in a second column; 3. an expected outcome in a third column; and 4. a next step in a fourth column. Preferably, a fifth column includes dependency conditions for the corresponding action step, where execution of the action step is conditional on the dependency condition being satisfied. The process can be executed by parsing a workflow program document represented by the table.

Automating complex processes with multiple interacting elements

Automating a process to process a task or an object involves defining elements of the process in one or more human-intelligible and editable and machine interpretable workflow program documents. The workflow program documents each include a plurality of actors who perform actions or take decisions, a sequence of action steps each associated with an actor and having at least one expected outcome and a corresponding next step for each expected outcome. The action steps are recorded in sequential form and have an option to include one or more dependencies on other actions or actors. The process is executed by a processor running the code defined by the workflow program documents. When executing the sequential actions, optional dependencies are evaluated and action steps are deferred notwithstanding their sequence in the documents if dependencies are not satisfied.

Abstract meaning representation parsing with graph translation

A computer-implemented method for generating an abstract meaning representation (“AMR”) of a sentence, comprising receiving, by a computing device, an input sentence and parsing the input sentence into one or more syntactic and/or semantic graphs. An input graph including a node set and an edge set is formed from the one or more syntactic and/or semantic graphs. Node representations are generated by natural language processing. The input graph is provided to a first neural network to provide an output graph having learned node representations aligned with the node representations in the input graph. The method further includes predicting via a second neural network, node label and predicting, via a third neural network, edge labels in the output graph. The AMR is generated based on the predicted node labels and predicted edge labels. A system and a non-transitory computer readable storage medium are also disclosed.

DETECTING SUITABILITY OF MACHINE LEARNING MODELS FOR DATASETS

Apparatuses, systems, program products, and method are disclosed for detecting suitability of machine learning models for datasets. An apparatus includes a training evaluation module configured to calculate a first statistical data signature for a training data set of a machine learning system using one or more predefined statistical algorithms. An apparatus includes an inference evaluation module configured to calculate a second statistical data signature for an inference data set of a machine learning system using one or more predefined statistical algorithms. An apparatus includes a score module configured to calculate a suitability score describing the suitability of a training data set to an inference data set as a function of a first and a second statistical data signature. An apparatus includes an action module configured to perform an action related to a machine learning system in response to a suitability score satisfying an unsuitability threshold.

Artificial intelligence system supporting semi-supervised learning with iterative stacking

A plurality of training iterations is conducted for a machine learning task. A given iteration includes generating a version of a stacking model using a portion of a labeled data set. Proposed labels are then obtained in the iteration using the generated version of the stacking model for a set of unlabeled records. The unlabeled records and their proposed labels are then used to generate versions of base models for the iteration. After the training iterations are completed, a trained ensemble of models including the stacking model and the base models is stored.

Automated digital tool identification from a rasterized image

A system identifies, automatically and without user intervention, digital tool parameters for achieving a visual appearance of an image region in raster image data. The system processes raster image data using a tool detection module 202 to identify the tool and a parameter estimation module 216 determines tool parameters for controlling the tool to achieve the visual appearance. The system generates an image tool description 110 based on the tool and the parameter configuration and incorporates the image tool description into the raster image data to create an interactive image 120 which indicates a region of the image having a visual appearance achievable using the tool. The tool detection and parameter estimation steps may be carried out by trained neural networks. The image tool description may enable transfer of the digital tool parameter configuration to different image data. Also claimed is a corresponding method and a system for generating training data used in training neural networks ...

Automating complex processes applicable to complex objects and tasks

Automated processing of a task or an object involves defining elements of the process in one or more human-intelligible and editable and machine interpretable workflow program documents 200. The documents each include a plurality of actors, a sequence of action steps each associated with an actor and having expected outcome(s) and a corresponding next step for each outcome. The action steps are recorded in sequential form and can include one or more dependencies on other actions or actors. Task definition document(s) record definitions of tasks or objects referenced in the workflow program documents using human-intelligible labels. The definition documents include parameters and attributes for each task or object. The process is executed by a processor running code defined by the workflow program documents. Sequential actions, tasks or objects are handled with reference to the labels. The workflow program documents are editable at a first user interface during execution of the process and ...

Training anonymized machine learning models via generalized data generated using received trained machine learning models

An example system includes a processor to receive training data and predictions on the training data of a trained machine learning model to be anonymized. The processor is to generate generalized data from training data based on the predictions of the trained machine learning model on the training data. The processor is to train an anonymized machine learning model using the generalized data.

Motion learning without labels

A machine learning model is described that is trained without labels to predict a motion field between a pair of images. The trained model can be applied to a distinguished pair of images to predict a motion field between the distinguished pair of images.

Method for identifying subsurface features

A method for improving a backpropagation-enabled process for identifying subsurface features from seismic data involves a model that has been trained with an initial set of training data. A target data set is used to compute a set of initial inferences on the target data set that are combined with the initial training data to define updated training data. The model is trained with the updated training data. Updated inferences on the target data set are then computed. A set of further-updated training data is defined by combining at least a portion of the initial set of training data and at least a portion of the target data and associated updated inferences. The set of further-updated training data is used to train the model. Further-updated inferences on the target data set are then computed and used to identify the occurrence of a user-selected subsurface feature in the target data set.

ANOMALOUS BEHAVIOR DETECTION

A training dataset is used to train an unsupervised machine learning trained model. Corresponding gradient values are determined for a plurality of entries included in the training dataset using the trained unsupervised machine learning model. A first subset of the training dataset is selected based on the determined corresponding gradient values and a first threshold value selected from a set of threshold values. A labeled version of the selected first subset is used to train a first supervised machine learning model to detect one or more anomalies.

Example-based image annotation

Techniques are generally described for machine learning exampled-based annotation of image data. In some examples, a first machine learning model may receive a query image comprising a first depiction of an object-of-interest. In some examples, the first machine learning model may receive a target image representing a scene in which a second depiction of the object-of-interest is visually represented. In various examples, the first machine learning model may generate annotated output image data that identifies a location of the second depiction of the object-of-interest within the target image. In some examples, an object detection model may be trained based at least in part on the annotated output image data.

Augmented reality enabled command management

The exemplary embodiments disclose a method, a computer program product, and a computer system for managing user commands. The exemplary embodiments may include a user giving one or more commands to one or more devices, collecting data of the one or more commands, extracting one or more features from the collected data, and determining which one or more of the commands should be executed on which one or more of the devices based on the extracted one or more features and one or more models.

NEURAL NETWORK TRACING ARRANGEMENTS

A sequence of images depicting an object is captured, e.g., by a camera at a point-of-sale terminal in a retail store. The object is identified, such as by a barcode or watermark that is detected from one or more of the images. Once the object's identity is known, such information is used in training a classifier (e.g., a machine learning system) to recognize the object from others of the captured images, including images that may be degraded by blur, inferior lighting, etc. In another arrangement, such degraded images are processed to identify feature points useful in fingerprint-based identification of the object. Feature points extracted from such degraded imagery aid in fingerprint-based recognition of objects under real life circumstances, as contrasted with feature points extracted from pristine imagery (e.g., digital files containing label artwork for such objects). A great variety of other features and arrangements—some involving designing classifiers so as to combat classifier ...

Communication efficient machine learning of data across multiple sites

In one embodiment, a service receives machine learning-based generative models from a plurality of distributed sites. Each generative model is trained locally at a site using unlabeled data observed at that site to generate synthetic unlabeled data that mimics the unlabeled data used to train the generative model. The service receives, from each of the distributed sites, a subset of labeled data observed at that site. The service uses the generative models to generate synthetic unlabeled data. The service trains a global machine learning-based model using the received subsets of labeled data received from the distributed sites and the synthetic unlabeled data generated by the generative models.

Method, device, and computer program product for self-supervised learning of pixel-wise anatomical embeddings in medical images

The present disclosure provides a method, a device, and a computer program product using a self-supervised anatomical embedding (SAM) method. The method includes randomly selecting a plurality of images; for each image of the plurality of images, performing random data augmentation to obtain a patch pair, generating global and local embedding tensors for each patch of the patch pair, and selecting positive pixel pairs from the patch pair and obtaining positive embedding pairs; for each positive pixel pair, computing global and local similarity maps, finding global hard negative embeddings, selecting global random negative embeddings, pooling the global hard negative embeddings and the global random negative embeddings to obtain final global negative embeddings, and finding local hard negative embeddings using the global and local similarity maps, and randomly sampling final local negative embeddings from the local hard negative embeddings; and minimizing a final info noise contrastive estimation ...

Facility level transaction-enabling systems and methods for provisioning and resource allocation

The present disclosure describes transaction-enabling systems and methods. A system can include a facility having a core task and a controller. The controller may include a facility description circuit to interpret historical facility parameter values and corresponding outcome values. A facility prediction circuit operates an adaptive learning system to train a facility resource allocation circuit in response to the historical facility parameter values and corresponding outcome values. The facility description circuit further interprets a plurality of present state facility parameter values and the trained facility resource allocation circuit adjusts facility resource values in response.

Processing structured and unstructured text to identify sensitive information

This disclosure describes techniques that include identifying sensitive information from any appropriate set of data, such as data produced by operations of a business or organization. In one example, this disclosure describes a method that includes receiving text data containing sensitive information, including structured sensitive information and unstructured sensitive information; applying a rule-based model to identify the structured sensitive information in the text data; applying a machine learning model to identify the unstructured sensitive information in the text data, wherein the machine learning model has been trained to identify unstructured sensitive information in text; and generating output text data from the text data by modifying the structured sensitive information identified by the rule-based model and the unstructured sensitive information identified by the machine learning model.

Extracting video clips from videos for use in training a machine-learning model

A video clip showing a wellsite activity can be extracted from a video and then labeled for use in training a machine-learning model in some examples described herein. In one such example, a system can train a model with a set of training data to identify an object and a corresponding spatial location of the object in each image frame in a video depicting a wellsite activity. The system can analyze image frames in the video using the trained model to identify a target image frame in which the object is present in a predefined spatial area thereof, where the predefined spatial area is associated with the wellsite activity. The system can then generate a video clip that includes only a subpart of the video based on the target image frame, where the subpart includes a series of consecutive image frames containing the target image frame.

Producing explainable rules via deep learning

A computer-implemented method according to one embodiment includes receiving, at a deep neural network (DNN), a plurality of sentences each having an associated label; training the DNN, utilizing the plurality of sentences and associated labels; and producing a linguistic expression (LE) utilizing the trained DNN.

Method of developing and using a tool and the resulting tool, for automating complex processes

Software for automating a process, for example a document handling process involving sending a contract for signatures. The process is defined in a tabular workflow program document, such as a spreadsheet. Each row represents a path through a portion of the process to an outcome. Each column holds a human readable label for a process component, such as a process portion name, an actor for carrying out that portion, an action to be carried out, an outcome, and a next step. The software filters the workflow program document to visualise and implement the workflow using an existing visual programming framework. The workflow is visualised as a flowchart using pre-built visual representations of workflow components. The workflow is implemented by programming associated with these visual representations. In this way, the workflow program document directs a portion of the process along one of the paths to an outcome based on the human readable labels. The software may be developed in a linear ...

METHOD AND SYSTEM FOR GENERATING A TRAINING DATASET

Forward market renewable energy credit prediction from automated agent behavioral data

Systems and methods for prediction of forward market renewable energy credit from automated agent behavioral data are disclosed. An example transaction-enabling system may include a forward market circuit to access a forward energy credit market, and a market forecasting circuit to automatically generate a forecast for a forward market price of an energy credit in the forward energy credit market, based in part on an automated agent behavior collected from an automated agent behavioral data source. The example system may further include wherein the energy credit comprises a renewable energy credit from a renewable energy system, and a smart contract circuit to sell the renewable energy credit or purchase the renewable energy credit on the forward energy credit market in response to the forecasted forward market price.

SCALABLE ARCHITECTURE FOR AUTOMATIC GENERATION OF CONTENT DISTRIBUTION IMAGES

Methods and systems are disclosed for automatic generation of content distribution images that include receiving user input corresponding to a content-distribution operation. The user input may be parsed to identify keywords. Image data corresponding to the keywords can be identified. Image-processing operations may be executed on the image data. Executing a generative adversarial network on the processed image data, which includes: executing a first neural network on the processed-image data to generate first images that correspond to the keywords, the first images generated based on a likelihood that each image of the first images would not be detected as having been generated by the first neural network. A user interface can display the first images with second images that include images that were previously part of content-distribution operations or images that were designated by an entity as being available for content-distribution operations.

Selecting a display with machine learning

Examples of methods performed by an electronic device are described. In some examples of the methods, a machine learning model is trained based on a plurality of interaction events and a corresponding plurality of images. In an example, each of the plurality of interaction events corresponds to one of a plurality of displays. In some examples of the methods, a display is selected of the plurality of displays based on the machine learning model. In an example, an object is presented on the display.

Semantic-aware feature engineering

In an embodiment, a process for semantic-aware feature engineering includes receiving semantic labels for data fields of training data. Each of the semantic labels is associated with a semantic meaning associated with a corresponding data field. The process includes automatically generating at least one new feature using at least a portion of the semantic labels.

PLATFORM FOR PERCEPTION SYSTEM DEVELOPMENT FOR AUTOMATED DRIVING SYSTEM

MACHINE LEARNING METHODS AND SYSTEMS FOR VARIETY PROFILE INDEX CROP CHARACTERIZATION

A system includes one or more processors; and one or more non-transitory, computer-readable media including instructions that, when executed by the one or more processors, cause the computing system to: receive a machine data set; process the machine data set with a trained deep learning model to generate predicted variety profile index values; and cause a visualization to be displayed.

SYSTEMS AND METHODS FOR CELL CLASSIFICATION

The present disclosure relates to automated systems and methods adapted to quickly and accurately train a neural network to detect and/or classify cells and/or nuclei. The present disclosure also relates to automated systems and methods for using a trained cell detection and classification engine, such as one including a neural network, to classify cells within an unlabeled image.

Apparatus and method of labeling for object detection

An apparatus of labeling for object detection according to an embodiment of the present disclosure includes an image selector that determines a plurality of labeling target images from among a plurality of unlabeled images, and determines a labeling order of the plurality of labeling target images, a feedback obtainer that obtains label inspection information on the plurality of labeling target images from a user, and a model trainer that learns the label inspection information input from the user by using the labeling target images, obtains a pseudo label for supervised learning based on a learning result using the label inspection information, and re-determines the labeling order of the labeling target images based on the pseudo label.

System and method for fashion attributes extraction

A system and a method for training an inference model using a computing device. The method includes: providing a text-to-vector converter; providing the inference model and pre-training the inference model using labeled fashion entries; providing non-labeled fashion entries; separating each of the non-labeled fashion entries into a target image and target text; converting the target text into a category vector and an attribute vector using the text-to-vector converter; processing the target image using the inference model to obtain processed target image and target image label; comparing the category vector to the target image label; when the category vector matches the target image label, updating the target image label based on the category vector and the attribute vector to obtain updated label; and retraining the inference model using the processed target image and the updated label.

SYSTEMS AND METHODS FOR SUBSCRIBER-BASED ADAPTATION OF PRODUCTION-IMPLEMENTED MACHINE LEARNING MODELS OF A SERVICE PROVIDER USING A TRAINING APPLICATION

A system and method for accelerating an adaptation of one or more machine learning models of a data handling and governance service includes implementing a data handling and governance platform digitally accessible by a target subscriber of the data handling and governance service, wherein the data handling and governance platform interfaces with a plurality of distinct subscriber-agnostic digital content machine learning classification models of the data handling and governance service; identifying a target subscriber-agnostic digital content machine learning classification model; adapting the target subscriber-agnostic digital content machine learning classification model to a subscriber-specific digital content machine learning classification model based on a training of the target subscriber-agnostic digital content machine learning classification model with at least one training corpus comprising subscriber-specific training data samples; and implementing the subscriber-specific digital ...

Adversarial network for transfer learning

Disclosed herein are arrangements that facilitate the transfer of knowledge from models for a source data-processing domain to models for a target data-processing domain. A convolutional neural network space for a source domain is factored into a first classification space and a first reconstruction space. The first classification space stores class information and the first reconstruction space stores domain-specific information. A convolutional neural network space for a target domain is factored into a second classification space and a second reconstruction space. The second classification space stores class information and the second reconstruction space stores domain-specific information. Distribution of the first classification space and the second classification space is aligned.

Optimizing training data for image classification

A method for machine learning-based classification may include training a machine learning model with a full training data set, the full training data set comprising a plurality of data points, to generate a first model state of the machine learning model, generating respective embeddings for the data points in the full training data set with the first model state of the machine learning model, applying a clustering algorithm to the respective embeddings to generate one or more clusters of the embeddings, identifying outlier embeddings from the one or more clusters of the embeddings, generating a reduced training data set comprising the full training data set less the data points associated with the outlier embeddings, training the machine learning model with the reduced training data set to a second model state, and applying the second model state to one or more data sets to classify the one or more data sets.

Apparatus and method for agricultural data collection and agricultural operations

Aspects of the subject disclosure may include, for example, obtaining video data from a single monocular camera, wherein the video data comprises a plurality of frames, wherein the camera is attached to a mobile robot that is travelling along a lane defined by a row of crops, wherein the row of crops comprises a first plant stem, and wherein the plurality of frames include a depiction of the first plant stem; obtaining robot velocity data from encoder(s), wherein the encoder(s) are attached to the robot; performing foreground extraction on each of the plurality of frames of the video data, wherein the foreground extraction results in a plurality of foreground images; and determining, based upon the plurality of foreground images and based upon the robot velocity data, an estimated width of the first plant stem. Additional embodiments are disclosed.

Machine learning methods and systems for variety profile index crop characterization

A system includes one or more processors; and one or more non-transitory, computer-readable media including instructions that, when executed by the one or more processors, cause the computing system to: receive a machine data set; process the machine data set with a trained machine-learned model to generate predicted variety profile index values; and cause a visualization to be displayed. A computer-implemented method includes receiving a machine data set; processing the machine data set with a trained machine-learned model to generate predicted variety profile index values; and causing a visualization to be displayed. A non-transitory computer-readable medium includes computer-executable instructions that, when executed by one or more processors, cause a computer to: receive a machine data set; process the machine data set with a trained machine-learned model to generate predicted variety profile index values; and cause a visualization to be displayed.

SYSTEMS AND METHODS FOR APPLYING A LENS FUNCTION GENERATED USING SUPERVISED LEARNING TECHNIQUES TO SEGMENT DATA PROVIDED TO AN UNSUPERVISED LEARNING MODEL

Creating text classification machine learning models

Techniques for creating a text classifier machine learning (ML) model are described. According to some embodiments, a language processing service finetunes a language ML model on unlabeled documents of a user, and then trains that finetuned language ML model on labeled documents of the user to be a text classifier that is customized for that user’s domain, e.g., the user’s documents. Additionally, the finetuned language ML model may be trained on labeled documents of the user, for prediction objectives for unlabeled data, before being trained as the text classifier.

Analysis apparatus, non-transitory computer-readable storage medium for analysis program, and analysis method

A method includes: generating a refine image having a maximized correct label score of inference from an incorrect image from which an incorrect label is inferred by a neural network; generating a third map by superimposing a first map and a second map, the first map indicating pixels to each of which a change is made in generating the refine image, of a plurality of pixels of the incorrect image, the second map indicating a degree of attention for each local region in the refine image, the each local region being a region that has drawn attention by the neural network; and specifying a set of pixels that cause incorrect inference in the incorrect image by calculating a pixel value of the third map for each set of pixels, wherein the map generating processing adjusts the second map based on appearance frequency of each degree of attention.

METHODS AND SYSTEMS FOR USING ARTIFICIAL INTELLIGENCE TO SELECT A COMPATIBLE ELEMENT

A system for using artificial intelligence to select a compatible element. The system includes at least a server wherein the at least a server is configured to receive training data. The at least a server is configured to receive at least a biological extraction from a user. The at least a server is configured to receive at least a datum of user activity data. The at least a server is configured to select at least a compatible element as a function of the training data, the at least a biological extraction, and the at least a user activity data. The at least a server is configured to transmit the at least a compatible element to a user client device.

DOMAIN ADAPTATION OF AI NLP ENCODERS WITH KNOWLEDGE DISTILLATION

Systems, methods, devices, instructions, and other examples are described for natural language processing. One example includes accessing natural language processing general encoder data, where the encoder data is generated from a general-domain dataset that is not domain specific. A domain specific dataset is accessed and filtered encoder data using a subset of the encoder data is generated. The filtered encoder data is trained using the domain specific dataset to generate distilled encoder data, and tuning values for the distilled encoder data are generated to configure task outputs associated with the domain specific dataset.

SYSTEMS AND METHODS FOR LABELING LARGE DATASETS OF PHYSIOLOGICAL RECORDS BASED ON UNSUPERVISED MACHINE LEARNING

A deep learning model and dimensionality reduction are applied to each of a plurality of records of physiological information to derive a plurality of feature vectors. A similarities algorithm is applied to the plurality of feature vectors to form a plurality of clusters, each including a set of feature vectors. An output comprising information that enables a display of one or more of the plurality of clusters is provided, and a mechanism for selecting at least one feature vector within a selected cluster of the plurality of clusters is enabled. Upon selection of a feature vector, an output comprising information that enables a display of the record of physiological information corresponding to the selected feature vector is provided, and a mechanism for assigning a label to the displayed record is enabled. The assigned label is then automatically assigned to the records corresponding to the remaining feature vectors in the selected cluster.

Transaction-enabled methods for providing provable access to a distributed ledger with a tokenized instruction set for a semiconductor fabrication process

Transaction-enabled methods for providing provable access to a distributed ledger with a tokenized instruction set for semiconductor fabrication processes are described. A method may include accessing a distributed ledger including an instruction set for a semiconductor fabrication process and tokenizing the instruction set. The method may further include interpreting an instruction set access request and, in response to the access request, providing a provable access to the instruction set. The method may further include providing commands to a production tool of the semiconductor fabrication process in response to the instruction set access request, and recording the transaction on the distributed ledger.

Image classification system, image classification method, and image classification program

An image classification system 10 includes: a probability computation means 11 which computes a known-image probability, which is the probability that an input image corresponds to a known image associated with a seen label that indicates the class into which content indicated by the known image is classified; a likelihood computation means 12 which computes both the likelihood that content indicated by the input image is classified into the same class as content indicated by an unseen image associated with an unseen label, and the likelihood that the content indicated by the input image is classified into the same class as the content indicated by the known image; and a correction means 13 which corrects each computed likelihood using the computed known-image probability.

Distance to obstacle detection in autonomous machine applications

In various examples, a deep neural network (DNN) is trained to accurately predict, in deployment, distances to objects and obstacles using image data alone. The DNN may be trained with ground truth data that is generated and encoded using sensor data from any number of depth predicting sensors, such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. Camera adaptation algorithms may be used in various embodiments to adapt the DNN for use with image data generated by cameras with varying parameters—such as varying fields of view. In some examples, a post-processing safety bounds operation may be executed on the predictions of the DNN to ensure that the predictions fall within a safety-permissible range.

Systems and methods for enabling machine resource transactions

The present disclosure describes transaction-enabling systems and methods for enabling machine resource transactions. A system can include a machine having at least one of a compute task requirement, a networking task requirement, and an energy consumption task requirement; and a controller. The controller can include a resource requirement circuit to determine an amount of a resource for the machine to service task requirement, a resource market circuit to access a resource market, and a resource distribution circuit to execute a transaction of the resource on the resource market in response to the determined amount of the resource.

Transaction-enabled methods for providing provable access to a distributed ledger with a tokenized instruction set for a biological production process

Transaction-enabled methods for providing provable access to a distributed ledger with a tokenized instruction set for biological production processes are described. A method may include accessing a distributed ledger comprising an instruction set for a biological production process and tokenizing the instruction set. The method may further include interpreting an instruction set access request and, in response to the access request, providing a provable access to the instruction set. The method may further include providing commands to a production tool of the biological production process in response to the instruction set access request and recording the transaction on the distributed ledger.

Method and apparatus for sample labeling, and method and apparatus for identifying damage classification

An embodiment provides a system and method for sample labeling. During operation, the system obtains a plurality of historical loss assessment images and obtains a plurality of candidate samples from the plurality of loss assessment images. A respective candidate sample comprises an image of a candidate damage area detected in a corresponding historical loss assessment image. The system clusters the plurality of candidate samples into a plurality of class clusters. For a respective class cluster, the system determines a center candidate sample set corresponding to a class cluster center of the respective class cluster, receives a manual labeling result associated with candidate samples in the determined center candidate sample set, and performs, according to the manual labeling result, damage classification labeling on other unlabeled candidate samples in the respective class cluster to obtain a plurality of labeled samples.

System and methods for data model detection and surveillance

A computer system is provided for monitoring and detecting changes in a data generating processes, which may be under a multi-dimensional and unsupervised setting. A target dataset is split into paired subgroups by a separator and one or more detectors are applied to detect changes, anomalies, inconsistencies, and the like between the paired subgroups. Metrics may be generated by the detector(s), which are then passed to an evaluating system.

Learning data generation apparatus, learning model generation system, learning data generation method, and non-transitory storage medium

A learning data generation apparatus includes an object extraction unit configured to extract an object image from an image; a classification evaluation unit configured to evaluate the object possireimage based on a learned model, and to calculate reliability indicating a degree of posibility that the object image is classified as a candidate label; a classification determination unit configured to, if the reliability is smaller than a first threshold and equal to or larger than a second threshold which is smaller than the first threshold, associate a temporary label different from the candidate label with the object image; and a learning data generation unit configured to generate learning data based on the object image that is associated with the temporary label.

Generating autonomous vehicle simulation data from logged data

Logged data from an autonomous vehicle is processed to generate augmented data. The augmented data describes an actor in an environment of the autonomous vehicle, the actor having an associated actor type and an actor motion behavior characteristic. The augmented data may be varied to create different sets of augmented data. The sets of augmented data can be used to create one or more simulation scenarios that in turn are used to produce machine learning models to control the operation of autonomous vehicles.

Depth data model training with upsampling, losses and loss balancing

Techniques for training a machine learned (ML) model to determine depth data based on image data are discussed herein. Training can use stereo image data and depth data (e.g., lidar data). A first (e.g., left) image can be input to a ML model, which can output predicted disparity and/or depth data. The predicted disparity data can be used with second image data (e.g., a right image) to reconstruct the first image. Differences between the first and reconstructed images can be used to determine a loss. Losses may include pixel, smoothing, structural similarity, and/or consistency losses. Further, differences between the depth data and the predicted depth data and/or differences between the predicted disparity data and the predicted depth data can be determined, and the ML model can be trained based on the various losses. Thus, the techniques can use self-supervised training and supervised training to train a ML model.

System and method for generating a training set for improving monocular object detection

Described are systems and methods for self-learned label refinement of a training set. In on example, a system includes a processor and a memory having a training set generation module that causes the processor to train a model using an image as an input to the model and 2D bounding based on 3D bounding boxes as ground truths, select a first subset from predicted 2D bounding boxes previously outputted by the model, retrain the model using the image as the input and the first subset as ground truths, select a second set of predicted 2D bounding boxes previously outputted by the model, and generate the training set by selecting the 3D bounding boxes from a master set of 3D bounding boxes that have corresponding 2D bounding boxes that form the second subset.

Systems and methods for machine forward energy and energy storage transactions

Systems and methods for machine forward energy and energy storage transactions are disclosed. An example transaction-enabling system may include a resource requirement circuit to aggregate a resource requirement for a fleet of machines to perform a task, wherein the resource requirement comprises an energy storage capacity requirement, a forward resource market circuit to access a forward market for energy, and a machine resource acquisition circuit to execute a transaction on the forward market for energy in response to the aggregated resource requirement.

ASSESSING PRESENT INTENTIONS OF AN ACTOR PERCEIVED BY AN AUTONOMOUS VEHICLE

Systems and methods for controlling an autonomous vehicle (AV). The methods comprise: generating candidate intentions of an actor based on a detected action of the actor and a classification associated with the actor; determining an overall probability for each candidate intention based on at least a persistence of the candidate intention over a non-interrupted sequence of cycles (where each cycle represents a time period over which the actor was sensed by a sensor); selecting candidate intention(s) based on the overall probabilities; forecasting a subsequent future intention that the actor may have after reaching a goal defined by the candidate intention(s) which was(were) selected; obtaining an actor trajectory that is consistent with the candidate intention(s) which was(were) selected and the subsequent future intention; and using the actor trajectory to influence a selected trajectory for AV.

System and Method for Training a Regression Neural Network for Localization of a Device in an Environment

The present disclosure provides a method and a system for training a neural network suitable for localization of a device within an environment based on signals received by the device. The method comprises training a bi-regressor neural network to identify locations from labeled data, wherein the bi-regressor neural network includes a feature extractor and a bi-regressor including two regressors; training parameters of the bi-regressor using the labeled data and unlabeled data, such that each of the two regressors identifies the same labeled locations while processing the labeled data and identifies different locations while processing the unlabeled data; and training parameters of the feature extractor using an adversarial discriminator to extract domain invariant features from the unlabeled data with statistical properties of the labeled data according to the adversarial discriminator such that each of the two regressors identifies the same locations while processing the domain invariant ...

DISTANCE TO OBSTACLE DETECTION IN AUTONOMOUS MACHINE APPLICATIONS

In various examples, a deep neural network (DNN) is trained to accurately predict, in deployment, distances to objects and obstacles using image data alone. The DNN may be trained with ground truth data that is generated and encoded using sensor data from any number of depth predicting sensors, such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. Camera adaptation algorithms may be used in various embodiments to adapt the DNN for use with image data generated by cameras with varying parameters—such as varying fields of view. In some examples, a post-processing safety bounds operation may be executed on the predictions of the DNN to ensure that the predictions fall within a safety-permissible range.

Video semantic segmentation method based on active learning

The present invention belongs to the technical field of computer vision, and provides a video semantic segmentation method based on active learning, comprising an image semantic segmentation module, a data selection module based on the active learning and a label propagation module. The image semantic segmentation module is responsible for segmenting image results and extracting high-level features required by the data selection module; the data selection module selects a data subset with rich information at an image level, and selects pixel blocks to be labeled at a pixel level; and the label propagation module realizes migration from image to video tasks and completes the segmentation result of a video quickly to obtain weakly-supervised data. The present invention can rapidly generate weakly-supervised data sets, reduce the cost of manufacture of the data and optimize the performance of a semantic segmentation network.

Systems and methods for machine learning models for performance measurement

Methods, systems, and computer-readable media for generating a statistically covaried machine learning model for performance measurement of service providers. The method receives a configuration file that includes one or more parameters associated with a plurality of individuals and parses it to generate and executing the database query on input data to generate sets of tabulated data of individuals of the plurality of individuals. The method next determines one or more measures of service providers listed in the configuration file using two or more tabulated data of individuals from the sets of tabulated data of individuals. The method finally generates a covaried machine learning model by training a machine learning model by statistically covarying measures and using them as training data.

SEMI-SUPERVISED KEYPOINT BASED MODELS

System and method for image segmentation

Methods and systems for image processing are provided. Image data may be obtained. The image data may include a plurality of voxels corresponding to a first plurality of ribs of an object. A first plurality of seed points may be identified for the first plurality of ribs. The first plurality of identified seed points may be labelled to obtain labelled seed points. A connected domain of a target rib of the first plurality of ribs may be determined based on at least one rib segmentation algorithm. A labelled target rib may be obtained by labelling, based on a hit-or-miss operation, the connected domain of the target rib, wherein the hit-or-miss operation may be performed using the labelled seed points to hit the connected domain of the target rib.

System for time-efficient assignment of data to ontological classes

Implementations are directed to receiving a set of training data including a plurality of data points, at least a portion of which are to be labeled for subsequent supervised training of a computer-executable machine learning (ML) model, providing at least one visualization based on the set of training data, the at least one visualization including a graphical representation of at least a portion of the set of training data, receiving user input associated with the at least one visualization, the user input indicating an action associated with a label assigned to a respective data point in the set of training data, executing a transformation on data points of the set of training data based on one or more heuristics representing the user input to provide labeled training data in a set of labeled training data, and transmitting the set of labeled training data for training the ML model.

SYSTEMS AND METHODS FOR FORWARD MARKET PURCHASE OF MACHINE RESOURCES

Systems and methods for automatically soliciting the purchase of a first or second machine-related resource in a forward market, wherein the first resource and the second resource are distinct instances of the same type of resource, are described. A sample system may include a fleet of machines, each having a resource requirement comprising at least two of: a compute resource, a spectrum resource, or a network bandwidth resource. The system may include an circuits to aggregate data corresponding to the machine-related resources from at least a behavioral data source, to determine a substitution cost of a second resource; to determine a machine-related resource acquisition value; and to automatically solicit a purchase, in a forward market, of one of the first resource or the second resource in response to the determined substitution cost of the second resource.

REGRESSION-BASED LINE DETECTION FOR AUTONOMOUS DRIVING MACHINES

In various examples, systems and methods are disclosed that preserve rich spatial information from an input resolution of a machine learning model to regress on lines in an input image. The machine learning model may be trained to predict, in deployment, distances for each pixel of the input image at an input resolution to a line pixel determined to correspond to a line in the input image. The machine learning model may further be trained to predict angles and label classes of the line. An embedding algorithm may be used to train the machine learning model to predict clusters of line pixels that each correspond to a respective line in the input image. In deployment, the predictions of the machine learning model may be used as an aid for understanding the surrounding environment—e.g., for updating a world model—in a variety of autonomous machine applications.

MACHINE LEARNING TO EXTRACT QUANTITATIVE BIOMARKERS FROM RF SPECTRUMS

The present disclosure provides for ultrasound systems and methods to pre-process ultrasound data to distinguish abnormal tissue from normal tissue. An exemplary method can include receiving a set of ultrasound data and partitioning the set into a set of windows. The method can then provide for processing the set of windows to determine a power spectrum for each window. The power spectrum for each window can be processed to determine a normalized power spectrum for each window. This normalized power spectrum can be processed for each window with a machine learning model. The method can then provide for displaying an image where each window of the set of windows is displayed using a unique identifier based on the output of the machine learning model.

Occupancy prediction neural networks

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating a future occupancy prediction for a region of an environment. In one aspect, a method comprises: receiving sensor data generated by a sensor system of a vehicle that characterizes an environment in a vicinity of the vehicle as of a current time point, wherein the sensor data comprises a plurality of sensor samples characterizing the environment that were each captured at different time points; processing a network input comprising the sensor data using a neural network to generate an occupancy prediction output for a region of the environment, wherein: the occupancy prediction output characterizes, for one or more future intervals of time after the current time point, a respective likelihood that the region of the environment will be occupied by an agent in the environment during the future interval of time.

Systems and methods for aggregating transactions and optimization data related to energy and energy credits

Systems and methods for aggregating transactions and optimization data related to energy and energy credits include a transaction-enabling system including a resource requirement circuit structured to aggregate a resource requirement for a fleet of machines to perform a task, wherein the resource requirement comprises an energy credit requirement; a forward resource market circuit structured to access a forward market for energy; and a machine resource acquisition circuit structured to execute a transaction on the forward market for energy in response to the aggregated resource requirement.

Scalable architecture for automatic generation of content distribution images

Methods and systems are disclosed for automatic generation of content distribution images that include receiving user input corresponding to a content-distribution operation. The user input may be parsed to identify keywords. Image data corresponding to the keywords can be identified. Image-processing operations may be executed on the image data. Executing a generative adversarial network on the processed image data, which includes: executing a first neural network on the processed-image data to generate first images that correspond to the keywords, the first images generated based on a likelihood that each image of the first images would not be detected as having been generated by the first neural network. A user interface can display the first images with second images that include images that were previously part of content-distribution operations or images that were designated by an entity as being available for content-distribution operations.

Systems and computer-implemented methods for identifying anomalies in an object and training methods therefor

A system identifies anomalies in an image of an object. An input image of the object containing zero or more anomalies is supplied to an image encoder. The image encoder generates an image model. The image model is applied to an image decoder that forms a substitute non-anomalous image of the object. Differences between the input image and the substitute non-anomalous image identify zero or more areas of the input image that contain the zero or more the anomalies. The system implements a flow-based model and has been trained using (a) a set of augmented anomaly-free images of the object applied at the image encoder and (b) a reconstruction loss calculated based on a norm of differences between each augmented anomaly-free image of the object and a corresponding output image from the image decoder.

SYSTEMS AND METHODS FOR PROCESSING ELECTRONIC IMAGES FOR COMPUTATIONAL DETECTION METHODS

Systems and methods are disclosed for receiving one or more electronic slide images associated with a tissue specimen, the tissue specimen being associated with a patient and/or medical case, partitioning a first slide image of the one or more electronic slide images into a plurality of tiles, detecting a plurality of tissue regions of the first slide image and/or plurality of tiles to generate a tissue mask, determining whether any of the plurality of tiles corresponds to non-tissue, removing any of the plurality of tiles that are determined to be non-tissue, determining a prediction, using a machine learning prediction model, for at least one label for the one or more electronic slide images, the machine learning prediction model having been generated by processing a plurality of training images, and outputting the prediction of the trained machine learning prediction model.

Classifier determination through label function creation and unsupervised learning

Software architectures relating to machine learning (e.g., relating to classifying sequential text data. Unlabeled sequential text data may be produced by a variety of sources such as text messages, email messages, message chats, social media applications, and web pages. Classifying such data may be difficult due to the freeform and unlabeled nature of text data from these sources. Thus, techniques for training a machine learning algorithm to classify unlabeled text data in freeform format. Training is based on generation of labelling functions from lexical databases, applying the labelling functions to unlabeled text data in an unsupervised manner, and generating trained classifiers that accurately classify the unlabeled text data. The trained classifiers may then be implemented classify text data accessed from the variety of sources. The present techniques provide high-quality and efficient labeling of unlabeled text data in freeform formats.

Image processing learning device, image processing learning method, and image processing learning program

Disclosed is a technology with which face direction estimation processing and face detection processing can be learned simultaneously and with high precision without incurring significant costs. The image processing learning device comprises: a face direction identification unit that identifies whether a face direction is already known; a position conversion unit that converts information regarding the face direction to a position on a manifold, when already known; a position estimation unit that estimates the position on the manifold, when unknown; a face identification unit that identifies whether an object is already known to be a face/not a face; a first update quantity calculation unit that calculates the update quantity according to whether the object is a face/not a face from the distance between the position on the manifold and the position in space, when already known; a second update quantity calculation unit that calculates the update quantity so as to be closer when the distance between the position on the manifold and the position in space is close, and further when far, when unknown; and a parameter update unit that updates parameters. The image processing learning device comprises: a face direction identification unit that identifies whether a face direction is already known; a position conversion unit that converts information regarding the face direction to a position on a manifold, when already known; a position estimation unit that estimates the position on the manifold, when unknown; a face identification unit that identifies whether an object is already known to be a face/not a face; a first update quantity calculation unit that calculates the update quantity according to whether the object is a face/not a face from the distance between the position on the manifold and the position in space, when already known; a second update quantity calculation unit that calculates the update quantity so as to be closer when the distance between the position ...

Using high definition maps for generating synthetic sensor data for autonomous vehicles

According to an aspect of an embodiment, operations may comprise accessing high definition (HD) map data of a region, presenting, via a user interface, information describing the HD map data, receiving instructions, via the user interface, for modifying the HD map data by adding one or more synthetic objects to locations in the HD map data, modifying the HD map data based on the received instructions, and generating a synthetic track in the modified HD map data comprising, for each of one or more vehicle poses, generated synthetic sensor data based on the one or more synthetic objects in the modified HD map data.

Automatic recognition of entities related to cloud incidents

Systems and methods for automatic recognition of entities related to cloud incidents are described. A method, implemented by at least one processor, for processing cloud incidents related information, including entity names and entity values associated with incidents having a potential to adversely impact products or services offered by a cloud service provider is provided. The method may include using at least one processor, processing the cloud incidents related information to convert at least words and symbols corresponding to a cloud incident into machine learning formatted data. The method may further include using a machine learning pipeline, processing at least a subset of the machine learning formatted data to recognize entity names and entity values associated with the cloud incident.

Distance estimation to objects and free-space boundaries in autonomous machine applications

In various examples, a deep neural network (DNN) is trained—using image data alone—to accurately predict distances to objects, obstacles, and/or a detected free-space boundary. The DNN may be trained with ground truth data that is generated using sensor data representative of motion of an ego-vehicle and/or sensor data from any number of depth predicting sensors—such as, without limitation, RADAR sensors, LIDAR sensors, and/or SONAR sensors. The DNN may be trained using two or more loss functions each corresponding to a particular portion of the environment that depth is predicted for, such that—in deployment—more accurate depth estimates for objects, obstacles, and/or the detected free-space boundary are computed by the DNN. In some embodiments, a sampling algorithm may be used to sample depth values corresponding to an input resolution of the DNN from a predicted depth map of the DNN at an output resolution of the DNN.

Systems and Methods for Identifying Unknown Instances

Systems and methods of the present disclosure provide an improved approach for open-set instance segmentation by identifying both known and unknown instances in an environment. For example, a method can include receiving sensor point cloud input data including a plurality of three-dimensional points. The method can include determining a feature embedding and at least one of an instance embedding, class embedding, and/or background embedding for each of the plurality of three-dimensional points. The method can include determining a first subset of points associated with one or more known instances within the environment based on the class embedding and the background embedding associated with each point in the plurality of points. The method can include determining a second subset of points associated with one or more unknown instances within the environment based on the first subset of points. The method can include segmenting the input data into known and unknown instances.

Driver attention monitoring method and apparatus and electronic device

Disclosed in the present disclosure are a driver attention monitoring method and apparatus and an electronic device. The method includes: capturing, by a camera arranged on a vehicle, a video of a driving area of the vehicle; determining, according to each of multiple frames of face images of a driver in the driving area included in the video, a type of a gazing area of the driver in the frame of face image, where the gazing area of each frame of face image is one of multiple types of defined gazing areas obtained by dividing a space area of the vehicle in advance; and determining an attention monitoring result of the driver according to a type distribution of gazing areas of the frames of face images included within at least one sliding time window in the video.

System and method for increasing data quality in a machine learning process

A method and system for increasing data quality of a dataset for semi-supervised machine learning analysis. The method includes: receiving known class label information for a portion of the data in the dataset; receiving clustering parameters from a user; determining a data cleanliness factor, and where the data cleanliness factor is below a predetermined cleanliness threshold: assigning data without class label information as a data point to a cluster using the clustering parameters, each cluster having a cluster class label associated with such cluster; and determining a measure of assignment, and where the measure of assignment for each data point is below a predetermined assignment threshold, receiving a class label for such data points, otherwise, assigning the respective cluster class label to each data point with the respective measure of assignment below the predetermined assignment threshold; and otherwise, outputting the dataset with associated class labels for machine learning analysis.

Superpixel classification method based on semi-supervised k-svd and multiscale sparse representation

The present invention discloses a superpixel classification method based on semi-supervised K-SVD and multiscale sparse representation. The method includes carrying out semi-supervised K-SVD dictionary learning on the training samples of a hyperspectral image; using the training samples and the overcomplete dictionary as the input to obtain the multiscale sparse solution of superpixels; and using the obtained sparse representation coefficient matrix and overcomplete dictionary to obtain the result of superpixel classification by residual method and superpixel voting mechanism. The proposing of the present invention is of great significance to solving the problem of salt and pepper noise and the problem of high dimension and small samples in the field of hyperspectral image classification, as well as the problem of how to effectively use space information in classification algorithm based on sparse representation.

System, method, and computer program for transformer neural networks

A system and method include one or more processing devices to implement a sequence of transformer neural networks, first and second sequence-to-sequence layers that each comprises a sequence of nodes, and an output layer to provide the first set and second set of score vectors to a downstream application of a natural language processing (NLP) task.

Model for mapping settlements

A programmable media includes a graphical processing unit in communication with a memory element. The graphical processing unit is configured to detect one or more settlement regions from a high resolution remote sensed image based on the execution of programming code. The graphical processing unit identifies one or more settlements through the execution of the programming code that executes a multi-instance learning algorithm that models portions of the high resolution remote sensed image. The identification is based on spectral bands transmitted by a satellite and on selected designations of the image patches.

Magnetic resonance imaging quality classification based on deep machine-learning to account for less training data

For classifying magnetic resonance image quality or training to classify magnetic resonance image quality, deep learning is used to learn features distinguishing between corrupt images base on simulation and measured similarity. The deep learning uses synthetic data without quality annotation, allowing a large set of training data. The deep-learned features are then used as input features for training a classifier using training data annotated with ground truth quality. A smaller training data set may be needed to train the classifier due to the use of features learned without the quality annotation.

Multi-source domain adaptation with mutual learning

In embodiments of the present disclosure, a method, device and computer-readable medium for multi-source domain adaptation are provided. The method comprises generating a first representation of a target image through a first trained classifier, generating a second representation of the target image through a second trained classifier, and generating a third representation of the target image through a third trained classifier. A mutual learning is conducted among the first, second and third classifiers during the training. The method further comprises determining a classification label of the target image based on the first, second and third representations. The present disclosure proposes a mutual learning network for multi-source domain adaptation, which can improve the accuracy of label generation for images.

Generative Adversarial Network Medical Image Generation for Training of a Classifier

Mechanisms are provided to implement a machine learning training model. The machine learning training model trains an image generator of a generative adversarial network (GAN) to generate medical images approximating actual medical images. The machine learning training model augments a set of training medical images to include one or more generated medical images generated by the image generator of the GAN. The machine learning training model trains a machine learning model based on the augmented set of training medical images to identify anomalies in medical images. The trained machine learning model is applied to new medical image inputs to classify the medical images as having an anomaly or not.

Information processing apparatus and information processing method

An information processing apparatus has an acquisitor configured to acquire an entire area image obtained by capturing an entire area of a processing surface of a wafer including at least one defect, a training image selector configured to select, as a training image, a partial image including at least one defect from the entire area image, a model constructor configured to construct a calculation model of generating a label image obtained by extracting and binarizing the defect included in the partial image, and a learner configured to update a parameter of the calculation model based on a difference between the label image generated by inputting the training image to the calculation model and a reference label image obtained by extracting and binarizing the defect of the training image.

Anomaly detection by classifying past behavior

Various embodiments relate to a method for detecting anomalies in a system by an anomaly detector, including: receiving a trained machine learning model that detects anomalies; receiving a set of new inputs from the to the anomaly detector from the system; setting a label for each of the set of new inputs to a value indicating normal operation of the system; training a new anomaly detection model using incremental learning to update the trained machine learning model using the labeled set of new inputs; receiving a set of past model inputs with an associated label; producing a verification set by inputting the set of past model inputs into the new anomaly detection model; and comparing the verification set with the labelled past model inputs to determine if an anomaly is present.

Image Enhancement Using Self-Examples and External Examples

Systems and methods are provided for image enhancement using self-examples in combination with external examples. In one embodiment, an image manipulation application receives an input image patch of an input image. The image manipulation application determines a first weight for an enhancement operation using self-examples and a second weight for an enhancement operation using external examples. The image manipulation application generates a first interim output image patch by applying the enhancement operation using self-examples to the input image patch and a second interim output image patch by applying the enhancement operation using external examples to the input image patch. The image manipulation application generates an output image patch by combining the first and second interim output image patches as modified using the first and second weights.

Creative gan generating music deviating from style norms

A method and system for generating music uses artificial intelligence to analyze existing musical compositions and then creates a musical composition that deviates from the learned styles. Known musical compositions created by humans are presented in digitized form along with a style designator to a computer for analysis, including recognition of musical elements and association of particular styles. A music generator generates a draft musical composition for similar analysis by the computer. The computer ranks such draft musical composition for correlation with known musical elements and known styles. The music generator modifies the draft musical composition using an iterative process until the resulting musical composition is recognizable as music but is distinctive in style.

Data augmentation for image classification tasks

A computer-implemented method and systems are provided for performing machine learning for an image classification task. The method includes overlaying, by a processor operatively coupled to one or more databases, a second image on a first image obtained from one or more training sets in the one or more databases, to form a mixed image, by averaging an intensity of each of a plurality of co-located pixel pairs in the first and the second image. The method also includes training, by the processor, a machine learning process configured for the image classification task using the mixed image to augment data used by the machine learning process for the image classification task.

Roof condition assessment using machine learning

Systems and methods for roof condition assessment from digital images using machine learning are disclosed, including receiving an image of a structure having roof characteristic(s), first pixel values depicting the structure, second pixel values outside of the structure depicting a background surrounding the structure, and first geolocation data; generating a synthetic shape image of the structure from the image using machine learning, including pixel values forming a synthetic outline shape, and having second geolocation data; mapping the synthetic shape onto the image, based on the first and second geolocation data, and changing the second pixel values so as to not depict the background; assessing roof characteristic(s) based on the first pixel values with a second machine learning algorithm resulting in a plurality of probabilities, each for a respective roof condition classification category, and determining a composite probability based upon the plurality of probabilities so as to classify the roof characteristic(s).

Method and apparatus for recommending sample data

The present disclosure proposes a method and an apparatus for recommending sample data. The method may include: inputting a plurality of pieces of sample data to be classified into at least one preset classification model, and acquiring a classifying probability of classifying each piece of sample data into each classification model; acquiring a first distance between each piece of sample data and a classifying boundary of each classification model according to the classifying probability of classifying the piece of sample data into the classification model, in which the classifying boundary of the classification model is configured to distinguish positive and negative sample data; computing a target distance for each piece of sample data according to the first distance between each piece of sample data and the classifying boundary of each classification model.

Method of training models in ai and electronic device

A method of training models in AI and an electronic device are disclosed, the electronic device is connected to other electronic devices and a controller, each electronic device is deployed with a single initial machine learning model and can obtain a prediction accuracy and weightings of neurons of the trained machine learning model. The controller determines new weightings from a plurality of the received weightings according to a preset rule and a plurality of received prediction accuracies. Each electronic device updates the weightings of neurons of the trained machine learning model to the new weightings. An electronic device is also disclosed. The method reduces a cost of training a machine learning model, utilizes network resources more efficiently, and improves an accuracy of the machine learning model.

Recognizing fine-grained objects in surveillance camera images

Systems and methods for recognizing fine-grained objects are provided. The system divides unlabeled training data from a target domain into two or more target subdomains using an attribute annotation. The system ranks the target subdomains based on a similarity to the source domain. The system applies multiple domain discriminators between each of the target subdomains and a mixture of the source domain and preceding target domains. The system recognizes, using the multiple domain discriminators for the target domain, fine-grained objects.

Transaction-enabled systems and methods for smart contracts

An example transaction-enabled system may include a smart contract wrapper to access a distributed ledger comprising intellectual property (IP) licensing terms corresponding to IP assets, wherein the IP licensing terms include an apportionment of royalties among owning entities in the distributed ledger. The smart contract wrapper may interpret an IP description value and an IP addition request, and, in response to the IP addition request and the IP description value, to add the apportionment of royalties corresponding to the IP description value. At least one of the plurality of IP assets comprises an instruction set and an operation on the distributed ledger provides provable access to the instruction set. A royalty apportionment wrapper apportions royalties from at least one royalty generating element to owning entities in response to the IP licensing terms.

Scalable semantic image retrieval with deep template matching

Approaches presented herein provide for semantic data matching, as may be useful for selecting data from a large unlabeled dataset to train a neural network. For an object detection use case, such a process can identify images within an unlabeled set even when an object of interest represents a relatively small portion of an image or there are many other objects in the image. A query image can be processed to extract image features or feature maps from only one or more regions of interest in that image, as may correspond to objects of interest. These features are compared with images in an unlabeled dataset, with similarity scores being calculated between the features of the region(s) of interest and individual images in the unlabeled set. One or more highest scored images can be selected as training images showing objects that are semantically similar to the object in the query image.

Method and system of building hospital-scale chest x-ray database for entity extraction and weakly-supervised classification and localization of common thorax diseases

A new chest X-ray database, referred to as “ChestX-ray8”, is disclosed herein, which comprises over 100,000 frontal view X-ray images of over 32,000 unique patients with the text-mined eight disease image labels (where each image can have multi-labels), from the associated radiological reports using natural language processing. We demonstrate that these commonly occurring thoracic diseases can be detected and spatially-located via a unified weakly supervised multi-label image classification and disease localization framework, which is validated using our disclosed dataset.

Systems and methods for forward market price prediction and sale of energy storage capacity

Systems and methods for forward market price prediction and sale of energy storage capacity are disclosed. An example transaction-enabling system may include a fleet of machines having an aggregate energy storage capacity; and a controller, comprising: an external data circuit structured to monitor an external data source and collect data from the external data source; an expert system circuit structured to predict a forward market price for energy storage capacity based on the collected data and the aggregate energy storage capacity; and a smart contract circuit structured to automatically sell at least a subset of the aggregate energy storage capacity on a forward market for energy storage capacity in response to the predicted forward market price.

Transaction-enabled systems and methods with licensing smart wrappers and ip aggregation

Transaction-enabled systems and methods with licensing smart wrappers and IP aggregation are disclosed. An example transaction-enabling system may include a smart wrapper structured to aggregate intellectual property (IP) assets into an aggregate stack of IP; and a smart contract wrapper configured to: access a distributed ledger comprising a plurality of intellectual property (IP) licensing terms corresponding to the aggregated IP; interpret an IP description value and an IP addition request; and in response to the IP addition request and the IP description value, add an IP asset to the aggregate stack of IP and commit a party to at least one of the plurality of IP licensing terms.

Transaction-enabled systems and methods for identifying and acquiring machine resources on a forward resource market

Transaction-enabled systems and methods for identifying and acquiring machine resources on a forward resource market are disclosed. An example system may include a controller having a resource requirement circuit to determine an amount of a resource required for a machine to service a task requirement, a forward resource market circuit to access a forward resource market, a resource market circuit to access a resource market, and a resource distribution circuit to execute a transaction of the resource on at least one of the resource market or the forward resource market in response to the determined amount of the resource required.

Edge-based adaptive machine learning for object recognition

Examples of techniques for adaptive object recognition for a target visual domain given a generic machine learning model are provided. According to one or more embodiments of the present invention, a computer-implemented method for adaptive object recognition for a target visual domain given a generic machine learning model includes creating, by a processing device, an adapted model and identifying classes of the target visual domain using the generic machine learning model. The method further includes creating, by the processing device, a domain-constrained machine learning model based at least in part on the generic machine learning model such that the domain-constrained machine learning model is restricted to recognize only the identified classes of the target visual domain. The method further includes computing, by the processing device, a recognition result based at least in part on combining predictions of the domain-constrained machine learning model and the adapted model.

Semi-Supervised Learning for Training an Ensemble of Deep Convolutional Neural Networks

The technology disclosed relates to constructing a convolutional neural network-based classifier for variant classification. In particular, it relates to training a convolutional neural network-based classifier on training data using a backpropagation-based gradient update technique that progressively match outputs of the convolutional network network-based classifier with corresponding ground truth labels. The convolutional neural network-based classifier comprises groups of residual blocks, each group of residual blocks is parameterized by a number of convolution filters in the residual blocks, a convolution window size of the residual blocks, and an atrous convolution rate of the residual blocks, the size of convolution window varies between groups of residual blocks, the atrous convolution rate varies between groups of residual blocks. The training data includes benign training examples and pathogenic training examples of translated sequence pairs generated from benign variants and pathogenic variants.

Systems and methods for aggregating transactions and optimization data related to energy credits

Systems and methods for aggregating transactions and optimization data related to energy credits are disclosed. An example transaction-enabling system may include a resource requirement circuit to aggregate an energy credit requirement for a fleet of machines to perform a task, a forward resource market circuit to access a forward market for energy, and a controller. The controller may include an artificial intelligence (AI) circuit to configure a transaction on the forward market for energy in response to the aggregated energy credit requirement; a machine resource acquisition circuit to automatically solicit the configured transaction on the forward market for energy; and wherein the AI circuit is further structured to iteratively improve the configured transaction to improve a task outcome of the fleet of machines.

Systems and methods for forward market energy price prediction and machine forward energy purchase

A transaction-enabling system includes a machine having an energy requirement for a task; and a controller, comprising: a resource requirement circuit structured to determine an amount of an energy resource for the machine to service the energy requirement; a market forecasting circuit structured to predict a forward market price on a forward resource market based on the determined amount of the energy resource and at least one external data source; a forward resource market circuit structured to access the forward resource market; and a resource distribution circuit structured to execute a transaction on the forward resource market in response to the determined amount of the energy resource and the predicted forward market price.

Techniques for dense video descriptions

Techniques and apparatus for generating dense natural language descriptions for video content are described. In one embodiment, for example, an apparatus may include at least one memory and logic, at least a portion of the logic comprised in hardware coupled to the at least one memory, the logic to receive a source video comprising a plurality of frames, determine a plurality of regions for each of the plurality of frames, generate at least one region-sequence connecting the determined plurality of regions, apply a language model to the at least one region-sequence to generate description information comprising a description of at least a portion of content of the source video. Other embodiments are described and claimed.

Image learning device, image learning method, neural network, and image classification device

In the image learning device according to an aspect of the invention, the neural network performs a first task of classifying a recognition target in a medical image and outputting a classification score as an evaluation result, and a second task different from the first task. The neural network updates a weight coefficient on the basis of a comparison result between the classification score output for the medical image of a first image group and a ground truth classification label, and does not reflect the classification score output for the medical image of a second image group in an update of the weight coefficient, for the first task. The neural network updates the weight coefficient on the basis of the evaluation result output for the medical image of the first image group and the evaluation result output for the medical image of the second image group, for the second task.

System and Method for Learning Random-Walk Label Propagation for Weakly-Supervised Semantic Segmentation

Systems and methods for training semantic segmentation. Embodiments of the present invention include predicting semantic labeling of each pixel in each of at least one training image using a semantic segmentation model. Further included is predicting semantic boundaries at boundary pixels of objects in the at least one training image using a semantic boundary model concurrently with predicting the semantic labeling. Also included is propagating sparse labels to every pixel in the at least one training image using the predicted semantic boundaries. Additionally, the embodiments include optimizing a loss function according the predicted semantic labeling and the propagated sparse labels to concurrently train the semantic segmentation model and the semantic boundary model to accurately and efficiently generate a learned semantic segmentation model from sparsely annotated training images.

Machine learning model training method and device, and expression image classification method and device

This application relates to a machine learning model training method and apparatus, and an expression image classification method and apparatus. The machine learning model training method includes: obtaining a machine learning model that includes a model parameter and that is obtained through training according to a general-purpose image training set; determining a sample of a special-purpose image and a corresponding classification label; inputting the sample of the special-purpose image to the machine learning model, to obtain an intermediate classification result; and adjusting the model parameter of the machine learning model according to a difference between the intermediate classification result and the classification label, continuing training, and ending the training in a case that a training stop condition is met. The solutions provided in this application improve the training efficiency of the machine learning model.

Systems and methods for self-learned label refinement for improving monocular object detection

Described are systems and methods for self-learned label refinement of a training set. In on example, a system includes a processor and a memory having a training set generation module that causes the processor to train a model using an image as an input to the model and 2D bounding based on 3D bounding boxes as ground truths, select a first subset from predicted 2D bounding boxes previously outputted by the model, retrain the model using the image as the input and the first subset as ground truths, select a second set of predicted 2D bounding boxes previously outputted by the model, and generate the training set by selecting the 3D bounding boxes from a master set of 3D bounding boxes that have corresponding 2D bounding boxes that form the second subset.

Evaluation process for a multi-task network

System, methods, and other embodiments described herein relate to evaluating a perception network in relation to the accuracy of depth estimates and object detections. In one embodiment, a method includes segmenting range data associated with an image according to bounding boxes of objects identified in the image to produce masked data. The method includes comparing the masked data with corresponding depth estimates in the depth map according to an evaluation mask that correlates the depth estimates with the depth map. The method includes providing a metric that quantifies the comparing to assess a network that generated the depth map and the bounding boxes.

Automatic labeling of products via expedited checkout system

A portable checkout unit automatically generates training data for an automatic checkout system as a customer collects items in a store. A customer uses an item scanner of portable checkout unit to generate a virtual shopping list of items collected in the shopping cart. When the customer adds a new item to the shopping cart or on some regular interval, the portable checkout unit captures images of the items contained by the shopping cart and can generate bounding boxes for each product in each image. The bounding boxes can be associated with item identifiers from previously-generated bounding boxes to identify the items captured by the bounding boxes. Each bounding box paired with an item identifier can then be used as training data for an automated checkout system.

Random and active learning for classifier training

An online system actively and randomly selects content items to be labeled for training a classifier. An online system receives content items from client devices of users and selects sets of the content items to be labeled by human labelers. The randomly selected content items are selected at random from the received content items, and the actively selected content items are selected based on the classifier's confidence in accurately predicting the classification of the content items. The online system may use a histogram of content items to actively select content items. The online system assigns the content items to bins of the histogram based on priority scores and selects content items with priority scores of the highest percentile. The online system provides the selected content items to human labelers for labeling. The labeled content items are then used for training the classifier.

Method and system for simultaneous scene parsing and model fusion for endoscopic and laparoscopic navigation

A method and system for scene parsing and model fusion in laparoscopic and endoscopic 2D/2.5D image data is disclosed. A current frame of an intra-operative image stream including a 2D image channel and a 2.5D depth channel is received. A 3D pre-operative model of a target organ segmented in pre-operative 3D medical image data is fused to the current frame of the intra-operative image stream. Semantic label information is propagated from the pre-operative 3D medical image data to each of a plurality of pixels in the current frame of the intra-operative image stream based on the fused pre-operative 3D model of the target organ, resulting in a rendered label map for the current frame of the intra-operative image stream. A semantic classifier is trained based on the rendered label map for the current frame of the intra-operative image stream.

Aligning unlabeled images to surrounding text

Aspects of the present invention disclose a method for extracting information of an unlabeled image within a document and aligning the information to text of the document. The method includes one or more processors identifying an image that is not associated with a corresponding label in a document that includes text. The method further includes determining a feature of an object of the image. The method further includes identifying an alignment candidate of the text of the document based at least in part on the feature of the object, wherein the alignment candidate is a segment of the text of the document identified as corresponding to the feature of the object. The method further includes aligning the feature with the alignment candidate of the text of the document.

Uncertainty guided semi-supervised neural network training for image classification

Aspects of the invention include systems and methods that train a teacher neural network using labeled images to obtain a trained teacher neural network, each pixel of each of the labeled images being assigned a label that indicates one of a set of classifications. A method includes providing a set of unlabeled images to the trained teacher neural network to generate a set of soft-labeled images, each pixel of each of the soft-labeled images being assigned a soft label that indicates one of the set of classifications and an uncertainty value associated with the soft label, and training a student neural network with a subset of the labeled images and the set of soft-labeled images to obtain a trained student neural network. Student-labeled images are obtained from unlabeled images using the trained student neural network.

System and method for predicting fall armyworm using weather and spatial dynamics

A dynamic graph includes a plurality of nodes and edges at a plurality of time steps; each node corresponds to a geographic location in a first area where pest infestation information is available for a subset of locations. Each edge connects two of the nodes which are geographically proximate, has a direction based on wind direction, and has a weight based on relative wind speed. Assign node features based on weather data as well as labels corresponding to pest infestation severity. Train a graph convolutional network on the dynamic graph. Based on predicted future weather conditions for a second area different than the first area, use the trained graph convolutional network to predict, via inductive learning, pest infestation severity for future times for a new set of nodes corresponding to new geographic locations in the second area for which no pest infestation information is available.

Learning systems and methods

A sequence of images depicting an object is captured, e.g., by a camera at a point-of-sale terminal in a retail store. The object is identified, such as by a barcode or watermark that is detected from one or more of the images. Once the object's identity is known, such information is used in training a classifier (e.g., a machine learning system) to recognize the object from others of the captured images, including images that may be degraded by blur, inferior lighting, etc. In another arrangement, such degraded images are processed to identify feature points useful in fingerprint-based identification of the object. Feature points extracted from such degraded imagery aid in fingerprint-based recognition of objects under real life circumstances, as contrasted with feature points extracted from pristine imagery (e.g., digital files containing label artwork for such objects). A great variety of other features and arrangements—some involving designing classifiers so as to combat classifier copying—are also detailed.

Apparatus, computer program product, and method for predictive data labelling using a dual-prediction model system

Various embodiments of the disclosure provide apparatuses, systems, and computer program products for predictive data labelling using a dual-model system. Embodiments provide various advantages in accuracy of predicted labels, for example in various contexts such as medical data analysis for difficult to diagnose diseases. An example provided apparatus is configured to generate a positive, neutral, and negative candidate identifier sets and corresponding positive, neutral, and negative candidate index sets based in part on applying a candidate selection rule set to a candidate data set; train a candidate label probabilistic model based at least in part on a candidate label training subset associated with the candidate data set associated with the positive and negative candidate identifiers; generate a candidate positive-label probability set using at least the candidate label probabilistic model; train a historical record prediction model to predict the candidate positive-label probability set; and utilize the historical record prediction model.

Multi-modal medical image processing

Aspects and/or embodiments seek to provide a method for training an encoder and/or classifier based on multimodal data inputs in order to classify regions of interest in medical images based on a single modality of data input source.

Method, apparatus, and system for generating synthetic image data for machine learning

An approach is provided for generating synthetic image data for machine learning. The approach, for instance, involves determining, by a processor, a set of parameters for indicating an action by one or more objects. The action is a dynamic movement of the one or more objects through a geographic space over a period of time. The approach also involves processing the set of parameters to generate synthetic image data. The synthetic image data includes a computer-generated image sequence of the one or more objects performing the action in the geographic space over the period of time. The approach further involves automatically labeling the synthetic image data with at least one label representing the action, the set of parameters, or a combination thereof. The approach further involves providing the labeled synthetic image data for training or evaluating a machine learning model to detect the action.

Method and system for improving quality of a dataset

A method and a server for updating a dynamic list of labeling tasks. One or more labels are received, each label associated to one labeling task; the one or more received labels are inserted into a dataset; an artificial intelligence (AI) model is trained on labeled data items from the dataset; predicted labels are obtained for a plurality of unlabeled data items from the dataset by applying the model thereon; a model-uncertainty measurement is computed by applying one or more regularization methods; relevancy values are computed for at least a subset of the predicted labels taking into account the predicted label and the model-uncertainty measurement; the data items corresponding to the labeling tasks with the highest relevancy values are inserted in the dynamic list; and the dynamic list is reordered upon computing of the relevancy values.

Systems and methods for managing organizational structures

Systems and methods are provided for managing organizational or corporate structures, including the employee roles or activities administered by human resources. A portion of the role datasets received within human resource records may be used to generate role tokens comprising unique datasets that have been truncated and deduped. Such tokens may be extracted based on assigned prioritization scores, and further assigned training labels representing categorical levels. Predictive labels may be assigned to a remaining portion of the extracted tokens via a logistic regression classifier, and a model organizational dataset may be generated based on the assigned training labels and the assigned predictive labels. The prediction certainty of the role tokens in the model organizational dataset may be used to map the identified role tokens to the roles represented in the human resource records.

Creating device, creating program, and creating method

To create a classifier whose classification accuracy is maintained in consideration of temporal changes in a generation distribution of a sample and a new feature that has not appeared in learning data, a classifier is created in which a feature correlation learning unit learns a correlation between a feature of a sample of labeled learning data, and a feature appearing only in a sample of unlabeled learning data, and a classifier creating unit adds the feature appearing only in the sample of the unlabeled learning data to the feature of the sample of the labeled learning data by using the correlation, and outputs a label associated with an input sample by using the sample of the labeled learning data to which the feature appearing only in the sample of the unlabeled learning data is added.

Explaining cross domain model predictions

Methods, systems, and computer program products for explaining cross domain model predictions are provided herein. A computer-implemented method includes providing a test data point to a domain adaptation model to obtain a prediction, wherein the domain adaptation model is trained on a set of labeled data points and a set of unlabeled data points. The method includes generating a task specific explanation for the prediction that includes one or more data points from among the sets that satisfy a threshold score for influencing the prediction. Additionally, the method includes generating a domain invariant explanation for the prediction. The domain variation explanation is generated by ranking pairs of data points based on a statistical similarity to the test data point, wherein each pair includes a data point from the set of labeled data points and a data point from the set of unlabeled data points.

Learning a lighting preference based on a reaction type

During operation, a computer provides, based at least in part on an initial lighting preference of an individual, instructions specifying initial lighting states of one or more lights in a lighting configuration in an environment, where an initial lighting state of a given light includes an intensity and a color of the given light. Then, the computer receives sensor data specifying a non-verbal physical response of the individual to initial lighting states. Moreover, the computer determines, based at least in part on the non-verbal physical response, a type of reaction of the individual to the initial lighting state. Next, the computer selectively modifies, based at least in part on a lighting behavior history of the individual and the determined type of reaction, the initial lighting preference of the individual to obtain an updated lighting preference.

Image signal provenance attestation

A computing device is configured to determine the provenance of an image. The computing device may receive an image. The computing device may generate an image capture profile associated with the image based at least in part on data generated during an image capture process. The computing device may determine whether the image is an authentic image based at least in part on the image capture profile. The computing device may, in response to determining that the image is an authentic image, generate a digital signature associated with the image.

Modification of Machine Learning Model Ensembles Based on User Feedback

Mechanisms are provided to implement an ensemble of unsupervised machine learning (ML) models. The ensemble of unsupervised ML models processes a portion of input data to generate an ensemble output and the ensemble output is output to an authorized user computing device to obtain user feedback from the authorized user via the user computing device. The user feedback indicates a correctness of the ensemble output. The mechanisms modify at least one feature of the ensemble of unsupervised ML models based on the obtained user feedback to thereby generate a modified ensemble of unsupervised ML models. Subsequent portions of input data are then processed using the modified ensemble of unsupervised ML models.

Gesture recognition method, gesture recognition system, and performing device therefore

A performing device of a gesture recognition system executes a performing procedure of a gesture recognition method. The performing procedure includes steps of: receiving a sensing signal; selecting one of sensing frames of the sensing signal; determining a soft label of the selected sensing frame; classifying a gesture event when the soft label of the selected sensing frame is approved. The gesture event is classified to determine the motion of the user. Therefore, the gesture recognition system does not need a predetermined time period to recognize the motion of the user. The time period for recognizing the motion of the user can be dynamical. A total time period for classifying a plurality of motions can be decreased, and the performance of the gesture recognition system can be improved.

Systems and methods for labeling large datasets of physiologial records based on unsupervised machine learning

A deep learning model and dimensionality reduction are applied to each of a plurality of records of physiological information to derive a plurality of feature vectors. A similarities algorithm is applied to the plurality of feature vectors to form a plurality of clusters, each including a set of feature vectors. An output comprising information that enables a display of one or more of the plurality of clusters is provided, and a mechanism for selecting at least one feature vector within a selected cluster of the plurality of clusters is enabled. Upon selection of a feature vector, an output comprising information that enables a display of the record of physiological information corresponding to the selected feature vector is provided, and a mechanism for assigning a label to the displayed record is enabled. The assigned label is then automatically assigned to the records corresponding to the remaining feature vectors in the selected cluster.

Deep Learning-Based Techniques for Training Deep Convolutional Neural Networks

The technology disclosed relates to constructing a convolutional neural network-based classifier for variant classification. In particular, it relates to training a convolutional neural network-based classifier on training data using a backpropagation-based gradient update technique that progressively match outputs of the convolutional neural network-based classifier with corresponding ground truth labels. The convolutional neural network-based classifier comprises groups of residual blocks, each group of residual blocks is parameterized by a number of convolution filters in the residual blocks, a convolution window size of the residual blocks, and an atrous convolution rate of the residual blocks, the size of convolution window varies between groups of residual blocks, the atrous convolution rate varies between groups of residual blocks. The training data includes benign training examples and pathogenic training examples of translated sequence pairs generated from benign variants and pathogenic variants.

Figure captioning system and related methods

Embodiments of the present invention are generally directed to generating figure captions for electronic figures, generating a training dataset to train a set of neural networks for generating figure captions, and training a set of neural networks employable to generate figure captions. A set of neural networks is trained with a training dataset having electronic figures and corresponding captions. Sequence-level training with reinforced learning techniques are employed to train the set of neural networks configured in an encoder-decoder with attention configuration. Provided with an electronic figure, the set of neural networks can encode the electronic figure based on various aspects detected from the electronic figure, resulting in the generation of associated label map(s), feature map(s), and relation map(s). The trained set of neural networks employs a set of attention mechanisms that facilitate the generation of accurate and meaningful figure captions corresponding to visible aspects of the electronic figure.

System and method for machine forward energy purchase based on model simulation on a digital twin

Systems and methods for machine forward energy purchase based on model simulation on a digital twin are disclosed. An example system may include an energy and compute facility including at least one of an energy source or an energy utilization requirement, and a controller. The controller may include a facility model circuit to operate a digital twin for the facility; a facility description circuit to interpret a set of parameters from the digital twin for the facility; and a facility configuration circuit to operate an adaptive learning system, wherein the adaptive learning system adjusts a facility configuration based on the set of parameters from the digital twin based, at least in part, on the energy source or the energy utilization requirement, and an energy credit forward market.

System and method for adjusting facility configuration based on model simulation on a digital twin

Systems and methods for machine forward energy purchase based on model simulation on a digital twin are disclosed. An example system may include an energy and compute facility including: at least one of an energy source or an energy utilization requirement; and a controller. The controller may include a facility model circuit to operate a digital twin for the facility; a facility description circuit to interpret a set of parameters from the digital twin for the facility; and a facility configuration circuit to operate an adaptive learning system, wherein the adaptive learning system is configured to adjust a facility configuration based on the set of parameters from the digital twin based at least in part on the energy source or the energy utilization requirement and an energy spot market.

Cyclic generative adversarial network for unsupervised cross-domain image generation

A system is provided for unsupervised cross-domain image generation relative to a first and second image domain that each include real images. A first generator generates synthetic images similar to real images in the second domain while including a semantic content of real images in the first domain. A second generator generates synthetic images similar to real images in the first domain while including a semantic content of real images in the second domain. A first discriminator discriminates real images in the first domain against synthetic images generated by the second generator. A second discriminator discriminates real images in the second domain against synthetic images generated by the first generator. The discriminators and generators are deep neural networks and respectively form a generative network and a discriminative network in a cyclic GAN framework configured to increase an error rate of the discriminative network to improve synthetic image quality.

Convolutional computing using multilayered analysis engine

Disclosed embodiments provide for deep convolutional neural network computing. The convolutional computing is accomplished using a multilayered analysis engine. The multilayered analysis engine includes a deep learning network using a convolutional neural network (CNN). The multilayered analysis engine is used to analyze multiple images in a supervised or unsupervised learning process. Multiple images are provided to the multilayered analysis engine, and the multilayered analysis engine is trained with those images. A subject image is then evaluated by the multilayered analysis engine. The evaluation is accomplished by analyzing pixels within the subject image to identify a facial portion and identifying a facial expression based on the facial portion. The results of the evaluation are output. The multilayered analysis engine is retrained using a second plurality of images.

Two-Stage Online Detection of Action Start In Untrimmed Videos

Embodiments described herein provide a two-stage online detection of action start system including a classification module and a localization module. The classification module generates a set of action scores corresponding to a first video frame from the video, based on the first video frame and video frames before the first video frames in the video. Each action score indicating a respective probability that the first video frame contains a respective action class. The localization module is coupled to the classification module for receiving the set of action scores from the classification module and generating an action-agnostic start probability that the first video frame contains an action start. A fusion component is coupled to the localization module and the localization module for generating, based on the set of action scores and the action-agnostic start probability, a set of action-specific start probabilities, each action-specific start probability corresponding to a start of an action belonging to the respective action class.

Systems and methods for automatic detection of an indication of abnormality in an anatomical image

There is provided a method for training a deep convolutional neural network (CNN) for detecting an indication of likelihood of abnormality, comprising: receiving anatomical training images, each including an associated annotation indicative of abnormality for the whole image without an indication of location of the abnormality, executing, for each anatomical training image: decomposing the anatomical training image into patches, computing a feature representation of each patch, computing for each patch, according to the feature representation of the patch, a probability that the patch includes an indication of abnormality, setting a probability indicative of likelihood of abnormality in the anatomical image according to the maximal probability value computed for one patch, and training a deep CNN for detecting an indication of likelihood of abnormality in a target anatomical image according to the patches of the anatomical training images, the one patch, and the probability set for each respective anatomical training image.

Image processing apparatus, image processing method, and image capture apparatus

An image processing apparatus that is capable of improving subject detection accuracy with respect to image signals is disclosed. The image processing apparatus applies subject detection processing to an image by using a learning model generated based on machine learning. The image processing apparatus selects the learning model from a plurality of learning models stored in advance, in accordance with characteristics of the image to which the subject detection processing is to be applied.

Scale inveriant object detection

Systems, and method and computer readable media that store instructions for scale invariant object detection.

Dynamic matching a sensed signal to a concept structure

A method for matching a sensed signal to a concept structure, the method may include: receiving a sensed signal; generating a signature of the input image; comparing the signature of the input image to signatures of a concept structure; determining whether the signature of the input image matches any of the signatures of the concept structure based on signature matching criteria, wherein each signature of the concept structure is associated within a signature matching criterion that is determined based on an object detection parameter of the signature; and concluding that the input image comprises an object associated with the concept structure based on an outcome of the determining.

Comprehensive Data Science Solution for Segmentation Analysis

A system, method, and computer-readable medium are disclosed for improved segmentation analysis. In various embodiments, an artificial learning blended algorithm (ALBA) system—is implemented. In various embodiments, the ALBA system includes an optimum cluster module to determine an optimum number of clusters for multiple clustering algorithms, and a validation cluster module to validate cluster algorithms using index validation techniques to determine a clustering algorithm from multiple clustering algorithms that use the determined optimum number of clusters.

Medical image assessment with classification uncertainty

Medical images may be classified by receiving a first medical image. The medical image may be applied to a machine learned classifier. The machine learned classifier may be trained on second medical images. A label of the medical image and a measure of uncertainty may be generated. The measure of uncertainty may be compared to a threshold. The first medical image and the label may be output when the measure of uncertainty is within the threshold.

System and method for data augmentation for trace dataset

A system and method for classification. In some embodiments, the method includes forming a first training dataset and a second training dataset from a labeled input dataset; training a first classifier with the first training dataset; training a variational auto encoder with the second training dataset, the variational auto encoder comprising an encoder and a decoder; generating a third dataset, by feeding pseudorandom vectors into the decoder; labeling the third dataset, using the first classifier, to form a third training dataset; forming a fourth training dataset based on the third dataset; and training a second classifier with the fourth training dataset.

Systems and Methods for Real-Time Adjustment of Neural Networks for Autonomous Tracking and Localization of Moving Subject

A goal of the disclosure is to provide real-time adjustment of a deep learning-based tracking system to track a moving individual without using a labeled set of training data. Disclosed are systems and methods for tracking a moving individual with an autonomous drone. Initialization video data of the specific individual is obtained. Based on the initialization video data, real-time training of an input neural network is performed to generate a detection neural network that uniquely corresponds to the specific individual. Real-time video monitoring data of the specific individual and the surrounding environment is captured. Using the detection neural network, target detection is performed on the real-time video monitoring data and a detection output corresponding to a location of the specific individual within a given frame of the real-time video monitoring data is generated. Based on the detection output, first tracking commands are generated to maneuver and center the camera on the location of the specific individual.

Counter rare training date for artificial intelligence

A system for enhancing a classifier prediction in respect to underrepresented classes may be provided. A classifier system trained with training data to build a model is used for classifying unknown input data, and an evaluator engine adapted for a determination of an underrepresented class. Additionally, the system comprises an extractor engine adapted for an extraction of relating data from an additional source, and a similarity engine adapted for a selection of data sets out of the relating data wherein the similarity engine is also adapted for comparing features of the relating data and a representative data set for the underrepresented class. Finally, the system comprises a recursion unit adapted for triggering the evaluator engine, the extractor engine and the similarity engine treating selected data set as input data until the evaluator engine classifies the selected data set with a confidence level which is above a confidence threshold level.

Calcium carbonate scale prediction and inhibition in hydrocarbon wells using machine learning

Methods for prediction and inhibition of calcium carbonate scale in hydrocarbon wells using machine learning include extracting training data including parameters from aqueous samples. Each aqueous sample is collected from a respective hydrocarbon well. The training data is classified in accordance with hydrocarbon production conditions of each hydrocarbon well. The classified training data is labeled in accordance with whether calcium carbonate scale has formed in each aqueous sample within a particular time period. A feature vector is determined from the labeled training data based on the parameters extracted from each aqueous sample. The feature vector is indicative of whether the respective hydrocarbon well contains calcium carbonate scale. A trained machine learning model is generated, wherein the machine learning model is trained based on the feature vector, to predict a number of the hydrocarbon wells containing calcium carbonate scale within the particular time period.

Adversarial Bootstrapping for Multi-Turn Dialogue Model Training

Systems described herein may use machine classifiers to perform a variety of natural language understanding tasks including, but not limited to multi-turn dialogue generation. Machine classifiers in accordance with aspects of the disclosure may model multi-turn dialogue as a one-to-many prediction task. The machine classifier may be trained using adversarial bootstrapping between a generator and a discriminator with multi-turn capabilities. The machine classifiers may be trained in both auto-regressive and traditional teacher-forcing modes, with the maximum likelihood loss of the auto-regressive outputs being weighted by the score from a metric-based discriminator model. The discriminators input may include a mixture of ground truth labels, the teacher-forcing outputs of the generator, and/or negative examples from the dataset. This mixture of input may allow for richer feedback on the autoregressive outputs of the generator. Additionally, dual sampling may improve response relevance and coherence by overcoming the problem of exposure bias.

Labeling using interactive assisted segmentation

Subject matter regards improving image segmentation or image annotation. A method can include receiving, through a user interface (UI), for each class label of class labels to be identified by the ML model and for a proper subset of pixels of the image data, data indicating respective pixels associated with the class label, partially training the ML model based on the received data, generating, using the partially trained ML model, pseudo-labels for each pixel of the image data for which a class label has not been received, and receiving, through the UT, a further class label that corrects a pseudo-label of the generated pseudo-labels.

Method for the improved detection of objects by a driver assistance system

The disclosure relates to a method for operating a driver assistance system of a motor vehicle. The method includes detecting a first data set of sensor data measured by a sensor device of the driver assistance program. The first data set of sensor data includes missing class allocation information, wherein the class allocation information relates to the objects represented by the sensor data. The method also includes pre-training a classification algorithm of the driver assistance system while taking into consideration the first data set in order to improve the object differentiation of the classification algorithm. The method further includes generating a second data set of simulated sensor data which includes at least one respective piece of class allocation information according to a specific specification. The method also includes training the classification algorithm of the driver assistance system while taking into consideration the second data set in order to improve an allocation assignment of the classification algorithm for objects differentiated by the classification algorithm. The method further includes improving the detection of objects, which are represented by additional measured sensor data, by the driver assistance system.

Updating learned models

Methods and systems are disclosed for updating learned models. An embodiment comprises receiving a plurality of data sets representing sensed data from one or more devices and determining, using one or more local learned models, local parameters based on the received data sets. Another operation may comprise generating a combined data set by combining the plurality of data sets and, determining, using one or more local learned models, global parameters based on the combined data set. Another operation may comprise transmitting, to a remote system, the global parameters for determining updated global parameters using one or more global learned models based at least partially on the global parameters, and receiving, from the remote system, the updated global parameters. Another operation may comprise updating the one or more local learned models using both the local parameters and updated global parameters.

Method and system for image segmentation and identification

An image segmentation method system, the system comprising: a training subsystem configured to train a segmentation machine learning model using annotated training data comprising images associated with respective segmentation annotations, so as to generate a trained segmentation machine learning model; a model evaluator; and a segmentation subsystem configured to perform segmentation of a structure or material in an image using the trained segmentation machine learning model. The model evaluator is configured to evaluate the segmentation machine learning model by (i) controlling the segmentation subsystem to segment at least one evaluation image associated with an existing segmentation annotation using the segmentation machine learning model and thereby generate a segmentation of the annotated evaluation image, and (ii) forming a comparison of the segmentation of the annotated evaluation image and the existing segmentation annotation. The method includes deploying the trained segmentation machine learning model for use if the comparison indicates that the segmentation machine learning model is satisfactory.

Method, system and computer program product for sentiment analysis

Methods, systems, and computer program product for automatically performing sentiment analysis on texts, such as telephone call transcripts and electronic written communications. Disclosed techniques include, inter alia, lexicon training, handling of negations and shifters, pruning of lexicons, confidence calculation for token orientation, supervised customization, lexicon mixing, and adaptive segmentation.

用于处理用于计算检测方法的电子图像的系统和方法

公开了用于接收与组织标本相关联的一个或多个电子载片图像的系统和方法，所述组织标本与患者和/或医学病例相关联，将所述一个或多个电子载片图像的第一载片图像划分成多个图块，检测所述第一载片图像和/或多个图块的多个组织区以生成组织掩模，确定所述多个图块中的任何图块是否对应于非组织，移除所述多个图块中被确定为非组织的任何图块，使用机器学习预测模型为所述一个或多个电子载片图像的至少一个标签确定预测，所述机器学习预测模型是通过处理多个训练图像而生成的，以及输出经训练的机器学习预测模型的预测。

基于半监督学习算法的非侵入式负荷监测方法及装置

本发明涉及一种基于半监督学习算法的非侵入式负荷监测方法及装置，其中方法包括：步骤S1：采集智能电表总用电量和各设备运行状态信息的时序信息；步骤S2，预处理数据，首先清洗数据，其次将数据归一化处理，最后以0填充总用电量序列的首尾；步骤S3，每次滑动一个时间长度来获得训练窗口，以总用电量序列作为输入窗口数据，以序列中点时间设备的开关状态作为输出标签，重复多次得到训练样本数据集；步骤S4，使用训练样本训练神经网络模型；步骤S5，将待识别的总电量序列输入训练好的神经网络模型，可以获得正确的各设备运行状态。与现有技术相比，本发明具有可以得到精细化的用户内部设备使用状态等优点。

A kind of lane line example detection method and apparatus

本发明公开一种车道线实例检测方法和装置，用以解决现有技术中无法准确有效地检测出车道线实例的问题。该方法包括：从车辆驾驶环境的图像数据中提取出车道线图像数据；其中，车道线图像中包括多个车道线像素；从车道线图像数据中识别出多组车道线像素；其中，每组内的车道线像素具有相同的特征；将一组车道线像素确定为一个车道线实例，得到多个车道线实例。

Based on the semi-supervised x-ray image automatic marking for generating confrontation network

本发明提出了一种基于生成对抗网络半监督X光自动标注方法，基于现有的生成对抗网络方法，改进了传统训练方法，利用监督损失和无监督损失相结合的半监督训练方法进行基于少量标注样本的图像分类识别。围绕X光图像标注数据稀缺性的问题进行研究，首先在传统无监督生成对抗网络基础上用softmax替换最后输出层，扩展成为半监督生成对抗网络，其次对生成样本定义额外类别标签引导训练，然后采用半监督训练对网络参数进行优化，最后采用训练好的判别网络对X光图像进行自动标注。该方法在医学X光图像自动标注方面，相比于传统监督学习和其他半监督学习算法性能得到了提高。

Rolling bearing fault diagnosis method under variable load based on unsupervised feature alignment

一种基于无监督特征对齐的变负载下滚动轴承故障诊断方法，属于滚动轴承故障诊断领域。针对滚动轴承实际工作中缺少某种负载数据，使得源领域数据与目标领域数据属于不同分布以及目标领域样本不含标签的问题。利用变分模态分解结合奇异值分解获取振动信号的时频特征，再结合振动信号时域、频域特征构建多域特征集；引入迁移学习中能够实现无监督领域适应的子空间对齐算法并进行改进，将核映射方法与SA算法相结合。将训练数据和测试数据映射到相同高维空间，在高维空间的子空间进行特征对齐，实现不同负载下源领域特征向目标领域特征对齐。在目标领域无标签的情况下，利用滚动轴承已知负载数据识别出其他负载数据对应的状态，具有较高的故障诊断准确率。

A kind of pedestrian's recognition methods again for learning and reordering based on unsupervised Local Metric

本发明公开了一种基于无监督局部度量学习和重排序的行人再识别方法，其实现过程为：首先获取行人图片，确定查询样本，并形成训练样本集和图片库；然后将获得的行人图片进行特征提取，并描述为特征向量；再为查询样本和/或图片库中各样本学习局部度量，得到对应的度量矩阵；最后通过学习得到的度量矩阵进行相似度计算，根据相似度大小进行初始排序；通过重排序优化初始排序得到最终的排序结果。本方法基于无监督局部度量学习，不需要人工标注样本，具有一定实用性和扩展性，通过重排序，进一步提高了匹配准确度。

Oil-well pump running condition recognition method and device

本发明提供了一种抽油泵运行状态识别方法及装置，本发明将目标灰度图像的目标特征向量输入至RWELM中，经RWELM运算后即可输出运行状态，以RWELM作为抽油泵运行状态识别的模型，以示功图和运行状态作为训练数据对RWELM进行训练，在训练过程中将结构风险最小化理论引入RWELM中，即采用调节参数对隐含层进行调节，解决了传统极限学习机过拟合问题，并且利用小波函数替代一般的隐含层激励函数，解决了极限学习机的局部最优的问题，具有诊断速度快、正确率高的优势。可以将本发明嵌入到抽油泵运行状态识别设备或系统中，能够及时发现运行状态与运行状态，为抽油泵的运行状态维修提供依据。

Multi-model handheld cloud detection transmission system and detection method based on 5G

本发明公开了一种基于5G的多模型手持云端检测传输系统及检测方法，基于5G的多模型手持云端检测传输系统，包括手持设备和云端检测模块，手持设备连接有5G网络且通过5G网络连接云端检测模块；手持设备内设有采集模块、处理模块、存储模块、检测模块和通信模块，采集模块连接处理模块，处理模块连接检测模块、存储模块和通信模块，检测模块连接存储模块和通信模块，通信模块连接5G网络。本发明通过手持设备通过5G网络实时接收学生网络检测模型进行检测，而云端检测模块则通过知识蒸馏模块进行教师网络检测模型的训练和更新，教师网络检测模型完成学生网络检测模型的训练和更新，可多设备同时工作，提高传输和处理效率。

Information extraction method, device, equipment and storage medium

本申请实施例公开了信息抽取方法、装置、设备以及存储介质，涉及图像处理技术领域。该方法的一具体实施方式包括：获取与目标文档影像的类别对应的位置模板；确定目标文档影像上的关键点位置；基于目标文档影像上的关键点位置和位置模板上的关键点位置，生成变换矩阵；基于位置模板上的信息位置和变换矩阵，确定目标文档影像对应的信息位置；对目标文档影像对应的信息位置处的信息进行抽取，得到目标文档影像中的信息。该实施方式通过构建特定类别的文档影像的位置模板，来确定该类别的文档影像对应的信息位置，从文档影像对应的信息位置处抽取信息，实现了简单、快速地信息抽取。

Method for determining visible mechanical continuous alarm of power transmission line channel

本发明公开一种输电线路通道可视化机械类连续告警的确定方法，基于已标注的输电线路通道可视化机械类连续告警样本数据，判断实时告警数据是否属于连续告警，具体涉及对某可视化巡视设备图像连续告警样本数据进行预处理、主成分降维分析提取数据特征和主成分矩阵、实时告警数据与主成分矩阵相乘运算获取对应特征、与提取的数据特征进行距离运算，通过阈值D 0 判断是否属于连续告警。本发明用特征数据对实时告警数据进行连续告警判定，解决了因实时告警数据变化多样，与单条连续告警样本比对漏报、误报率高的问题，为后续的应用场景如告警等级智能标注、AI图像识别模型疑似误报及漏报样本识别提供了模型支撑，进而提高输电线路运检的智能化水平。

A method, a device, an electronic equipment and a storage medium for extracting information

본 출원의 실시예는 이미지 처리 기술 분야에 관한 것으로 정보 추출 방법, 장치, 기기 및 저장 매체를 개시한다. 상기 정보 추출 방법은, 대상 문서 이미지의 카테고리에 대응되는 위치 템플릿을 획득하는 단계; 대상 문서 이미지 상에서의 제1 키포인트 위치를 결정하는 단계; 제1 키포인트 위치 및 위치 템플릿 상에서의 제2 키포인트 위치에 기반하여, 변환 행렬을 생성하는 단계; 위치 템플릿 상에서의 정보 위치 및 변환 행렬에 기반하여, 대상 문서 이미지에 대응되는 정보 위치를 결정하는 단계; 및 대상 문서 이미지에 대응되는 정보 위치에 위치한 정보를 추출하여, 대상 문서 이미지에 포함된 정보를 획득하는 단계를 포함한다. 상기 실시형태는, 특정 카테고리의 문서 이미지의 위치 템플릿을 생성하여, 상기 카테고리의 문서 이미지에 대응되는 정보 위치를 결정하고, 문서 이미지에 대응되는 정보 위치에서 해당 정보를 추출함으로써, 간단하고 신속한 정보 추출을 구현한다.

User intention recognition method, device, electronic apparatus, computer readable storage media and computer program

【課題】大量の訓練サンプルを自動的に選別し、さらに訓練して精度の高い意図認識モデルを得て、意図認識精度を向上させる。【解決手段】ユーザ意図認識方法は、複数回の対話データ及び毎回の対話データの満足度を取得するとともに、満足度が所定の満足条件を満たす対象対話データを選別し、対象対話データにおける入力データに意図ラベルを付け、対象対話データにおける入力データ及び入力データの意図ラベルに基づいて意図認識モデルを訓練して、訓練済み意図認識モデルを通じて新たな入力データに対し意図認識を行う。【選択図】図１

Cutting fluid amount adjustment device and cutting fluid amount adjustment system

Hyperspectral classification method based on expanding morphology and Active Learning

本发明公开一种基于扩展形态学与主动学习的高光谱图像分类方法，解决现有技术不能充分挖掘高光谱图像空间信息，导致分类精度低的问题。其步骤为：1)输入高光谱图像数据；2)对数据降维，提取光谱特征，并通过形态学剖面变换，得到空间特征；3)融合空谱特征，划分训练与测试样本集；4)利用训练样本集进行SVM分类；5)主动学习循环，由MCLU准则和AP聚类选取样本标记，更新训练与测试样本集；6)利用新的训练样本集进行SVM分类，直到训练样本数量达到预设数量时停止，得到最终分类结果。本发明将多结构元素的形态学特征与主动学习相结合，充分利用空谱信息，在小样本前提下提高了分类精度。

A kind of High Resolution SAR image classification method based on depth convolution ladder network

本发明公开了一种基于深度卷积阶梯网络的高分辨SAR图像分类方法，输入待分类的高分辨SAR图像及其标记信息；构造训练数据集D1与测试数据集D2；对数据集D1、D2的特征进行归一化得到数据集D3、D4；构造基于深度卷积阶梯网络的分类器模型；用训练数据集D3对网络进行训练；利用训练好的分类模型对测试数据集D4进行分类。本发明可充分利用少量有类标的训练样本，且采用卷积层有效提取高层判别特征而获得较高的分类精度。

The Method and System for Evaluating the Quality of Medical Image dataset for Machine Learning

본 발명은 의료 영상 데이터가 기계학습에 이용하기 적합한지 여부를 확인하기 위하여 의료 영상 데이터 세트의 품질을 평가하는 방법 및 그 시스템에 관한 것으로, 평가항목은 전체 프레임 중 정상 프레임이 차지하는 비율을 의미하는 데이터 정상성; 수신한 데이터에서 레이블링 되거나 레이블링 가능한 프레임이 차지하는 비율을 의미하는 학습 적합성; 및 의료 표준에 기초한 해부학적 요소에 대비하여 수신한 데이터에 포함된 해부학적 요소의 비율을 의미하는 해부학적 완전성을 포함할 수 있다.

Hyperspectral image robust classification method based on segmented depth features and low-rank representation

本发明提供了一种基于分段深度特征及低秩表示的高光谱图像鲁棒分类方法。首先，为了尽可能降低噪声对特征提取的影响，使用基于堆栈去噪自编码器网络对高光谱图像进行无监督地特征提取；然后，通过充分挖掘高光谱图像中类内的相似性及类间的差异性，建立基于低秩表示的鲁棒分类器；最后，采用有效的优化方法对目标函数进行优化求解。在训练数据中存在噪声的情况下，也能获得较好的分类效果。

Systems and methods for training prediction system for depth perception

【課題】境界ボックスを使用して奥行き予測システムを訓練するシステム及び方法を提供すること。【解決手段】一実施形態では、本方法は、画像をセグメント化することにより、境界ボックスを超えた領域をマスクして、境界ボックスの内部の非マスク領域を識別することを含む。該方法は、また、非マスク領域内で画像の画素と関連付けられた重み付け点をグランドトゥルース奥行きと比較することからの奥行き損失を使用して、奥行きモデルを訓練することを含む。該方法は、また、物体検出のために奥行きモデルを提供することを含む。【選択図】図５

Model training method, device, electronic equipment and computer readable storage medium

本申请涉及一种模型训练方法、装置、电子设备和计算机可读存储介质。上述方法包括：A:将图像数据输入预设的深度学习模型，获取所述深度学习模型对所述图像数据进行检测得到的第一检测信息；B:获取对所述第一检测信息进行校正得到的第二检测信息；C:根据所述图像数据及对应的第二检测信息对所述深度学习模型进行训练；D:获取训练后深度学习模型的收敛结果；E:在所述收敛结果不满足预设收敛条件时，迭代执行步骤A至步骤D直到所述收敛结果满足预设收敛条件。上述方法，采用深度学习模型标注的图像数据作为训练的训练数据，对深度学习模型进行训练的训练数据集无需人工标注，降低了人工标注数据训练集的工作量，节省了训练深度学习模型的成本。

Object tracking method and apparatus for a non-overlapping-sensor network