Abstract: Enclosed are embodiments for scoring one or more trajectories of a vehicle through a given traffic scenario using a machine learning model that predicts reasonableness scores for the trajectories. In an embodiment, human annotators, referred to as a “reasonable crowd,” are presented with renderings of two or more vehicle trajectories traversing through the same or different traffic scenarios. The annotators are asked to indicate their preference for one trajectory over the other(s). Inputs collected from the human annotators are used to train the machine learning model to predict reasonableness scores for one or more trajectories for a given traffic scenario. These predicted trajectories can be used to rank trajectories generated by a route planner based on their scores, compare AV software stacks, or used by any other application that could benefit from a machine learning model that scores vehicle trajectories.

Abstract: System and method of deploying a trained machine learning neural network (MLNN) in generating a fall injury condition of a subject. The method comprises receiving, at input layers of the trained MLNN, millimeter wave (mmWave) radar point cloud data representing fall attributes from monitoring the subject via mmWave radar sensing device, the input layers associated with the fall attributes, receiving, at a second set of input layers, personal attributes of the subject associated with ones of the second set of input layers, the first and second sets of input layers interconnected with an output layer of the trained MLNN via intermediate layers, the trained MLNN produced by establishing a correlation between an injury condition of prior subjects and mmWave point cloud data and personal attributes associated with the prior subjects, and generating, at the output layer, the fall injury condition attributable to the subject.

Abstract: The present disclosure provides a data processing method, an apparatus, an electronic device and a medium, which relates to the technical fields of autonomous driving, electronic maps, deep learning, image processing, and the like. The method includes: a computing device inputs a reference image and a captured image into a feature extraction model; obtain, a set of reference descriptors based on the first descriptor map; determine a plurality of sets of training descriptors; obtain a predicted pose of the vehicle by inputting the plurality of training poses and a plurality of similarities into a pose prediction model; and train the feature extraction model and the pose prediction model. When applied to a vehicle localization system, the trained feature extraction model and pose prediction model according to some embodiments of the present disclosure can improve accuracy and robustness of vehicle localization, thereby boosting the performance of the vehicle localization system.

Abstract: Generating visual data by defining a first action into a first set of objects and corresponding first set of motions, and defining a second action into a second set of objects and corresponding second set of motions. A relationship is then determined for the second action to the first action in terms of relationships between corresponding constituent objects and motions. Objects and motions are detected from visual data of first action. Visual data is composed for the second action from the data by transforming the constituent objects and motions detected in first action based on the corresponding determined relationships.

Abstract: An apparatus for training an image classification model according to an embodiment disclosed includes a first trainer that trains a model body and a first head through supervised learning based on a labeled data set subjected to type 1 labeling, a second trainer that trains the model body, the first head, and a second head through multi-task learning based on the labeled data set and an unlabeled data set, and a third trainer that trains a plurality of third heads through supervised learning based on the labeled data set subjected to type 2 labeling while freezing the model body.

Abstract: A computer implemented method for algorithmically editing digital video content is disclosed. A video file containing source video is processed to extract metadata. Label taxonomies are applied to extracted metadata. The labelled metadata is processed to identify higher-level labels. Identified higher-level labels are stored as additional metadata associated with the video file. A clip generating algorithm applies the stored metadata for selectively editing the source video to generate a plurality of different candidate video clips. Responsive to determining a clip presentation trigger on a viewer device, a clip selection algorithm is implemented that applies engagement data and metadata for the candidate video clips to select one of the stored candidate video clips. The engagement data is representative of one or more engagement metrics recorded for at least one of the stored candidate video clips. The selected video clip is presented to one or more viewers via corresponding viewer devices.

Abstract: According to the present disclosure, a weight parameter of a neural network is divided into a plurality of portions having a certain size and approximation is individually performed on the portions using a weighted sum of the codebook vectors.

Abstract: Automatic object optimization to accelerate machine learning training is disclosed. A request for a machine learning training dataset comprising a plurality of objects is received from a requestor. The plurality of objects includes data for training a machine learning model. A uniqueness characteristic for objects of the plurality of objects is determined, the uniqueness characteristic being indicative of how unique each object is relative to each other object. A group of objects from the plurality of objects is sent to the requestor, the group of objects being selected based at least partially on the uniqueness characteristic or sent in an order based at least partially on the uniqueness characteristic.

Abstract: Systems and methods for analyzing image data to identify cabinet products are disclosed. A computer-implemented method may include receiving, from an electronic device via a network connection, at least one digital image depicting a cabinet. The method also may include analyzing, by one or more processors, the at least one digital image to determine a first set of characteristics of the cabinet. Additionally, the method may include accessing, by the one or more processors from memory, a second set of characteristics corresponding to a plurality of cabinet products and comparing the first set of characteristics to the second set of characteristics to identify a cabinet product of the plurality of cabinet products that matches the cabinet. Further, the method may include transmitting, to the electronic device via the network connection, an indication of the cabinet product.

Abstract: A computer-implemented method of training an autoencoder to recover missing data is provided. The autoencoder includes an encoder for encoding its inputs into a latent space and a decoder for decoding the encodings from the latent space. The method comprises creating a first training set including a valid data set of multiple dimensions, and training the encoder and the decoder in a first training stage using the first training set to reduce a difference between the valid data set provided to the encoder and a data set decoded by the decoder. The method further comprises creating a second training set comprising an invalid data set, and training the encoder in a second training stage using the second training set to reduce a difference between encodings of valid data instances and encodings of their corresponding invalid data instances.

Abstract: An information processing apparatus includes a hardware processor which (i) performs learning by a learning data set associated with a correct answer label for a preset problem and creates a machine learning model for estimating a correct answer to the preset problem for input data, (ii) estimates the correct answer to the preset problem for the input data by using the machine learning model, (iii) in response to a user operation, determines a label indicating a result of the estimation as a correct answer label of the input data or corrects the label to determine the corrected label as a correct answer label of the input data, and (iv) additionally registers the determined correct answer label as learning data in association with the input data in the learning data set.

Abstract: A system for auto classification of products includes an entity recognizer and a model selector. The entity recognizer receives training data including an attribute of a product. The model selector selects a feature from the training data using a first statistical model to provide a first feature and a second statistical model to provide a second feature, and trains a probabilistic classifier using the first and the second features for providing a first and a second classification models respectively. Further, the model selector calculates an accuracy score of the obtained classification models for each distinct category in a preset hierarchy of categories and selects a classification model from the obtained classification models based on the accuracy score. The selected classification model has a highest accuracy score for a corresponding category in the preset hierarchy.

Abstract: A method of generating log data that includes a plurality of records, in which information detected through image recognition processing on a plurality of image frames is recorded. The method includes, detecting objects in each image frame through the image recognition processing, generating a record including identification information of each image frame and the aggregated number of objects detected in the image frame for each image frame, and adding metadata relating to the detected objects to the corresponding record when the aggregated number of objects is equal to or greater than one. When the aggregated number of objects is zero, the metadata related to the detected objects is not added to the record.

Abstract: A method for identifying a scene, comprising a computing device receiving a plurality of data points corresponding to a scene; the computing device determining one or more subsets of data points from the plurality of data points that are indicative of at least one sub-scene in said scene, said at least one sub-scene displayed on a display device that is part of said scene, wherein said at least one sub-scene does not represent said scene; the computing device categorizing said scene, disregarding said at least one sub-scene, wherein the categorizing includes interpreting said scene by a computer vision system such that said at least one sub-scene is not taken into account in the categorizing of said scene.

Abstract: A computer-implemented method, a computer program product, and a system for reducing labeled sample quantities required to update test sets. The computer-implemented method includes inputting a portion of unlabeled production data into a base model and generating labeled output relating to the unlabeled production data. The computer-implemented method also includes inputting the labeled output into a performance predictor. The performance predictor is a meta model of the base model that is trained with another portion of the unlabeled production data, a training set used to train the base model, and a test set portioned from the training set. The computer-implemented method further includes outputting, by the performance predictor, a performance metric relating to the labeled output produced by the trained base model. The performance metric can be any metric capable of measuring the output performance of the base model.

Abstract: A method of selecting an artificial intelligence (AI) model based on input data to select an AI model capable of correctly obtaining a result corresponding to data to be classified, e.g., a classification result indicating one of at least one class, from among a plurality of AI models, and a display device for performing the method.

Abstract: A system and method includes generating approximate distributions for distinct classes of data samples; computing a first partial Jensen-Shannon (JS) divergence and a second partial JS divergence based on the approximate distribution of the disparity affected class of data samples with reference to the approximate distribution of the control class of data samples; computing a disparity divergence based on the first partial JS divergence and the second partial JS divergence; generating a distribution-matching term based on the disparity divergence, wherein the distribution-matching term mitigates an inferential disparity between the control class of data samples and the disparity affected class of data samples during a training of an unconstrained artificial neural network; constructing a disparity-constrained loss function based on augmenting a target loss function with the distribution-matching term; and transforming the unconstrained ANN to a disparity-constrained ANN based on a training of the unconstraine

Abstract: An image processing apparatus 10 includes an interface 12 configured to acquire an image and a processor 13 configured to perform low-rank approximation by singular value decomposition on the acquired image, and perform object recognition on the acquired image subjected to the low-rank approximation by the singular value decomposition.

Abstract: Systems and methods for automatically detecting, classifying, and processing objects captured in an images or videos are provided. In one embodiment, the system receives an image from an image source and detects one or more objects in the image. The system performs a high-level classification of the one or more objects in the image. The system performs a specific classification of the one or more objects, determines a price of the one or more objects, and generates a pricing report comprising a price of the one or more objects. In another embodiment, the system captures at least one image or video frame and classifies an object present in the image or video frame using a neural network. The system adds the classified object and an assigned object code to an inventory and processes the inventory to assign the classified object a price.

Abstract: A measuring system 1 includes a server 200 identifying a kind of a product from a product image in which the product is included and a measuring device 100 identifying the kind of the product from the target image in which the product is included. The server 200 includes an acquisition unit that acquires a product image and product information relating to a kind of a product, a dividing unit that acquires a plurality of divided imaged by dividing the product image into a plurality of areas, and a generation unit that generates an identifying model by performing machine learning on the basis of a plurality of divided images extracted by an extraction unit that extracts a plurality of divided images satisfying a predetermined condition relating to a shown amount of the product from among the plurality of divided images.

Abstract: A supervised training device and method for training a neural network, and a supervised classification method and device based on the neural network are provided. The training device includes a storage unit and a processor. The processor accesses and executes a sampling module, a labelling module and a training module in the storage unit. The sampling module samples a first image and a second image from a first dataset. The labelling module tags the first image in response to a first control instruction to generate a first tagged image, and generates label data according to the first tagged image and the second image. The training module trains the neural network according to the label data.

Abstract: A method and system learn functions to be associated with data fields of forms to be incorporated into an electronic document preparation system. The functions are essentially sets of operations required to calculate the data field. The method and system receive form data related to a data field that expects data values resulting from performing specific operations. The method and system utilize machine learning and training set data to generate, test, and evaluate candidate functions to determine acceptable functions.

Abstract: Provided are a learning data collection apparatus, a learning data collection method, and a program for collecting learning data to be used for efficient retraining.

Abstract: Provided are an image processing apparatus and an image processing method that process a far-infrared image. The image processing apparatus includes a region extraction section, a modal transformation section, and a superimposition section. The region extraction section extracts a region of interest within a visible-light image captured by a visible-light camera. The modal transformation section receives an image of the region of interest within an infrared image captured by an infrared camera observing the same subject as the visible-light camera, and transforms the received image to a modal image. The superimposition section generates a presentation image by superimposing the modal image on the region of interest within the visible-light image. The modal transformation section transforms a far-infrared image of the region of interest to a modal image including an information modal familiar to humans by using, for example, a database and a conditional probability distribution.

Abstract: Systems and methods for performing spatial line grouping on digital ink stokes. The system includes an electronic processor configured to access a set of hypothetical lines in an electronic document and determine a set of hypothetical line pairings. The electronic processor is also configured to determine, via a gradient boosting tree model, a merge confidence score for each hypothetical line pairing and compare a first merge confidence score with a merge threshold. The first merge confidence score is associated with a first hypothetical line and a first neighboring hypothetical line. The electronic processor is also configured to, in response to the first merge confidence score satisfying the merge threshold, merge the first hypothetical line and the first neighboring hypothetical line to form a first line grouping. The electronic processor is also configured to perform a digital ink stroke analysis on the electronic document based on the first line grouping.

Abstract: A measuring system 1 includes a server 200 identifying a kind of a product from a product image in which the product is included and a measuring device 100 identifying the kind of the product from the target image in which the product is included. The server 200 includes an acquisition unit that acquires a product image and product information relating to a kind of a product, a dividing unit that acquires a plurality of divided imaged by dividing the product image into a plurality of areas, and a generation unit that generates an identifying model by performing machine learning on the basis of a plurality of divided images extracted by an extraction unit that extracts a plurality of divided images satisfying a predetermined condition relating to a shown amount of the product from among the plurality of divided images.

Abstract: The current embodiments relate to a real-time automated classification system that uses machine learning system to recognize important moments in broadcast content based on log data and/or other data received from various classification systems. The real-time automated classification system may be trained to recognize correlations between the various log data to determine key moments in the broadcast content. The real-time automated logging system may determine and generate metadata that describe or give information about what is happening or appearing in the broadcast content. The real-time automated logging system may automatically generate control inputs, suggestions, recommendations, and/or edits relating to broadcast content based upon the metadata, during broadcasting of the broadcast content.

Abstract: In various embodiments, a computer-implemented method of training a neural network for relighting an image is described. A first training set that includes source images and a target illumination embedding is generated, the source images having respective illuminated subjects. A second training set that includes augmented images and the target illumination embedding is generated, where the augmented images corresponding to the source images. A first autoencoder is trained using the first training set to generate a first output set that includes estimated source illumination embeddings and first reconstructed images that correspond to the source images, the reconstructed images having respective subjects that are i) from the corresponding source image, and ii) illuminated based on the target illumination embedding.

Abstract: Provided are a method and an apparatus for performing scalable video decoding, wherein the method and the apparatus down-sample input video, determine the down-sampled input video as base layer video, generate prediction video for enhancement layer video by applying an up-scaling filter to the base layer video, and code the base layer video and the prediction video, wherein the up-scaling filter is a convolution filter of a deep neural network.

Abstract: The present application relates to a semiconductor memory training method and related devices, belonging to the technical field of semiconductors. The method comprises: obtaining a stored historical training result of a semiconductor memory, the historical training result comprising a historical expected delay value and a historical expected voltage; setting a delay threshold and a current training voltage range, the delay threshold being less than or equal to the historical expected delay value, the current training voltage range comprising the historical expected voltage; obtaining a current minimum delay value for the semiconductor memory under the historical expected voltage; and using the stored historical training result as a current training result of the semiconductor memory, if the current minimum delay value for the semiconductor memory under the historical expected voltage is no less than the delay threshold.

Abstract: Systems, methods, and computer-readable for multi-spatial scale object detection include generating one or more object trackers for tracking at least one object detected from on one or more images. One or more blobs are generated for the at least one object based on tracking motion associated with the at least one object. One or more tracklets are generated for the at least one object based on associating the one or more object trackers and the one or more blobs, the one or more tracklets including one or more scales of object tracking data for the at least one object. One or more uncertainty metrics are generated using the one or more object trackers and an embedding of the one or more tracklets. A training module for detecting and tracking the at least one object using the embedding and the one or more uncertainty metrics is generated using deep learning techniques.

Abstract: A global and local binary pattern image crack segmentation method based on robot vision comprises the following steps: enhancing a contrast of an acquired original image to obtain an enhanced map; using an improved local binary pattern detection algorithm to process the enhanced map and construct a saliency map; using the enhanced map and the saliency map to segment cracks and obtaining a global and local binary pattern automatic crack segmentation method; and evaluating performance of the obtained global and local binary pattern automatic crack segmentation method. The present application uses logarithmic transformation to enhance the contrast of a crack image, so that information of dark parts of the cracks is richer. Texture features of a rotation invariant local binary pattern are improved. Global information of four directions is integrated, and the law of universal gravitation and gray and roundness features are introduced to correct crack segmentation results, thereby improving segmentation accuracy.

Abstract: The present approach relates to a system capable of life-long learning in a deep learning context. The system includes a deep learning network configured to process an input dataset and perform one or more tasks from among a first set of tasks. As an example, the deep learning network may be part of an imaging system, such as a medical imaging system, or may be used in industrial applications. The system further includes a learning unit communicatively coupled to the deep learning network 102 and configured to modify the deep learning network so as to enable it to perform one or more tasks in a second task list without losing the ability to perform the tasks from the first list.

Abstract: A system and methods for presenting a hybrid cloud cache as a file system. The system implements a set of standard file system command line interfaces that present the objects stored by the hybrid cloud cache to users of the system as if the users were viewing and interacting with a traditional file system. The system provides an interactive shell to the users to view the contents of the hybrid cloud cache. The system may be configured to operate on a live instance as well as on an on-disk structure of the hybrid cloud cache. The system may provide the ability to present partially cached cloud data as a file system via the interactive shell for the purposes of development, support, and troubleshooting.

Abstract: Aspects relate to systems and methods for parsing and correlating solicitation video content. An exemplary system includes a computing device configured to receive a solicitation video related to a subject, where the solicitation video includes at least an image component and at least an audio component, where the audio component includes audible verbal content related to at least an attribute of the subject, transcribe at least a keyword as a function of the audio component, and associate the subject with at least a job description as a function of the at least a keyword.

Abstract: Training a machine learning neural network (MLNN) in radiowave based monitoring of fall characteristics in diagnosing injury. The method comprises receiving, in a first set of input layers of the MLNN, from a millimeter wave (mmWave) radar sensing device, a set of mmWave radar point cloud data representing fall attributes associated with a subject, each of the first set associated with a respective fall attribute; receiving, at a second set of input layers of the MLNN, a set of personal attributes of the subject, training a MLNN classifier based on supervised training that establishes a correlation between an injury condition of the subject as generated at the output layer, the mmWave point cloud data, and personal attributes; and adjusting an initial matrix of weights by backpropagation to increase correlation between the injury condition, the mmWave point cloud data, and the personal attributes.

Abstract: An object identification system on a mobile work machine receives an object detection sensor signal from an object detection sensor, along with an environmental sensor signal from an environmental sensor. An object identification system generates a first object identification based on the object detection sensor signal and the environmental sensor signal. Object behavior is analyzed to determine whether the object behavior is consistent with the object identification, given the environment. If an anomaly is detected, meaning that the object behavior is not consistent with the object identification, given the environment, then a secondary object identification system is invoked to perform another object identification based on the object detection sensor signal and the environmental sensor signal. A control signal generator can generate control signals to control a controllable subsystem of the mobile work machine based on the object identification or the secondary object identification.

Abstract: In one embodiment, a vehicle battery diagnostics system forecasts a future state for a battery by monitoring, over a period of time, one or more of voltage, current or temperature signals from at least one battery of the vehicle, storing information from the voltage, current or temperature signals as time-series data, obtaining a forecasting model from a server, the forecasting model indicating at least one shapelet feature that corresponds to a forecast categorization, identifying, in the time-series data, a shapelet that matches the at least one shapelet feature to a degree exceeding a predetermined similarity threshold, and providing a notification indicating the forecast categorization.

Abstract: Systems and methods to provide a recommendation for an action based on an application policy are disclosed. The application policy may be associated with an organization. An application policy engine can use an artificial intelligence (AI) engine to execute a machine learning (ML) model. The application policy engine may receive real time video data or audio data, and obtain metadata comprising reference data or environment data. The application policy engine can process the real time video data or audio data using the ML model to infer biometric characteristics associated with a subject. The application policy engine can determine if the application policy was met, conformed to, or missed based on a correlation between the metadata and the inferred biometric characteristics, and provide a corresponding recommendation for an action.

Abstract: The present invention discloses a method and a system for monitoring manufacturing operation workflow using Structural Similarity (SSIM) index based activity detection. The method comprising receiving video data corresponding to a manufacturing operation activity, extracting a plurality of video frames from the video data, measuring SSIM index for each video frame of the plurality of video frames with respect to next consecutive video frame of the plurality of video frames, comparing the SSIM index of the each video frame with the SSIM index of next consecutive video frame of the plurality of video frames to identify one or more local maxima, and determining at least one manufacturing operation activity based on the one or more local maxima using machine learning technique.

Abstract: 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.

Abstract: A display apparatus is provided. The display apparatus includes an input interface, a first storage, a display, and a processor. Pixel values corresponding to a predetermined number of lines in an image input through the input interface are stored in the first storage. The processor acquires a first patch of a predetermined size by sampling a number of pixel values located in an outer region of a matrix centering about a specific pixel value from among the pixel values stored in the first storage, acquires a high-frequency component for the specific pixel value based on the acquired first patch, and processes the input image based on the high-frequency component. The display displays the processed image.

Abstract: A method is described for determining a confidence value for an object of a class determined by a neural network in an input image. The method includes: preparing an activation signature with the aid of a multiplicity of output images of a layer of the neural network for the class of the object, with the input image being provided to the input of the neural network; scaling the activation signature to the size of the input image; comparing an overlapping area portion of an area of the activation signature with an area of an object frame in relation to the area of the activation signature in order to determine the confidence value.

Abstract: Provided are a detecting device, a detecting method, a generating method, and a computer-readable storage medium that allow the user to readily obtain information on the degree of wear for a worn portion in the human-powered vehicle. A detecting device includes a control unit that detects a worn portion in a human-powered vehicle as a target worn portion from a first image including at least a part of the human-powered vehicle and outputs wear information related to a degree of wear for the target worn portion.

Abstract: Methods and systems for automated faux-manual image-marking of a digital image are disclosed, including a method comprising obtaining results of an automated analysis of one or more digital image indicative of determinations of structure abnormalities of one or more portions of a structure depicted in the one or more digital image; applying automatically, on the one or more digital image, with one or more computer processors, standardized markings indicative of the location in the image of the structure abnormalities of the structure depicted in the image; and generating, automatically with the one or more computer processors, one or more faux-manual markings by modifying one or more of the standardized markings, utilizing one or more image-manipulation algorithm, wherein the faux-manual markings mimic an appearance of manual markings on the structure in the real world.

Abstract: Techniques are described for inpainting of image data with a missing region. In an embodiment, at each iteration, the process determines a corresponding missing boundary region of the missing region and generates a collection of boundary patches for the missing boundary region. Based on comparing a boundary patch from the collection to source patches from a known source region of image data, the process generates replacement patches for the missing boundary region. When a boundary pixel data unit corresponds to multiple replacement pixel data units from different replacement patches, the process aggregates the multiple replacement pixel data units to generate an updated boundary pixel data unit. In an embodiment, the process performs convolution using the updated and previously known region of the image data.

Abstract: A method of performing image retrieval includes: obtaining a query image; generating a global feature descriptor of the query image by inputting the query image into a convolutional neural network (CNN) and obtaining the global feature descriptor as an output of the CNN, where parameters of the CNN are learned during training of the CNN on a batch of training images using a listwise ranking loss function and optimizing a quantized mean average precision ranking evaluation metric; determining similarities between the query image and other images based on distances between the global feature descriptor of the query image and global feature descriptors of the other images, respectively; ranking the other images based on the similarities, respectively; and selecting a set of the other images based on the similarities between the query image and the other images.

Abstract: This patent concerns novel technology for detecting backdoors of neural network, particularly deep neural network (DNN), classifiers. The backdoors are planted by suitably poisoning the training dataset, i.e., a data-poisoning attack. Once added to input samples from a source class (or source classes), the backdoor pattern causes the decision of the neural network to change to a target class. The backdoors under consideration are small in norm so as to be imperceptible to a human, but this does not limit their location, support or manner of incorporation. There may not be components (edges, nodes) of the DNN which are dedicated to achieving the backdoor function. Moreover, the training dataset used to learn the classifier may not be available.

Abstract: A system, method, and computer-readable storage medium are disclosed that execute machine vision operations to categorize a locality. At least one embodiment accesses a map image of a locality, where the map image includes geographical artefacts corresponding to entities within the locality; analyzes the map image to detect the entities in the locality using the geographical artefacts; assigns entity classes to detected entities in the locality; assigns a locality score to the locality based on entity classes included in the locality; retrieves street view images for one or more of the detected entities in the locality; and analyzes street view images of the detected entities to assign one or more further classifications to the detected entities. Other embodiments include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the method.

Abstract: A diagnostic imaging support system includes: an image input unit that receives an input of an image; a first processing unit that detects a region of interest including a subject of interest from the image, classifies the detected region of interest, and outputs a first classification result; and a second processing unit that subjects the region of interest to semantic segmentation, classifies the region of interest based on a result of the semantic segmentation, and outputs a second classification result. When a predetermined condition related to the region of interest is met, the first classification result is employed as indicating a class of the region of interest, and, when the condition is not met, the second classification result is employed as indicating the class of the region of interest.