Abstract: An electronic device for image correction and a method therefore are provided. The electronic device includes a camera, a memory, a communication circuit, and at least one processor. The at least one processor is configured to obtain a raw image of an external object using the camera, determine whether correction of the raw image is necessary, when the correction of the raw image is determined not to be necessary, store an image corresponding to the raw image in the memory, and when the correction of the raw image is determined to be necessary, transmit the raw image to an external electronic device via the communication circuit so as to enable the external electronic device to generate a corrected image of the raw image, and store the corrected image in the memory.

Abstract: Systems and methods for iris authentication are disclosed. In one aspect, a deep neural network (DNN) with a triplet network architecture can be trained to learn an embedding (e.g., another DNN) that maps from the higher dimensional eye image space to a lower dimensional embedding space. The DNN can be trained with segmented iris images or images of the periocular region of the eye (including the eye and portions around the eye such as eyelids, eyebrows, eyelashes, and skin surrounding the eye). With the triplet network architecture, an embedding space representation (ESR) of a person's eye image can be closer to the ESRs of the person's other eye images than it is to the ESR of another person's eye image. In another aspect, to authenticate a user as an authorized user, an ESR of the user's eye image can be sufficiently close to an ESR of the authorized user's eye image.

Abstract: Techniques for real-time location tagging are disclosed. A camera-enabled device presents an image of a physical space currently being captured by the camera-enabled device. Responsive to detecting user input while the image of the physical space is being captured by the camera-enabled device, the camera-enabled device tags a location in a plan of the physical space with information based at least in part on the user input.

Abstract: There is provided a control method for an image processing apparatus that generates a plurality of pieces of album data corresponding to a plurality of albums, the plurality of pieces of album data including a common page, which is a page common to all of the plurality of pieces of album data, and an individual page, which is a page not common to all of the plurality of pieces of album data. At least one image data group out of at least one image data group specified as a common image data group is controlled in such a way as not to be specified as an individual image data group.

Abstract: There is provided a method for image segmentation and annotation. The method comprises: providing an image; reducing the image size and smoothing the image to perform low-pass filtering on the image; restoring the image size; performing pixel level clustering of the image; performing spatial clustering of the image; selecting one or more clusters corresponding to a region of interest, ROI, based on a predetermined image mask, wherein the predetermined image mask has a known location in relation to the ROI; defining the selected one or more clusters as a proposed ROI; and determining if the proposed ROI sufficiently corresponds to an actual ROI.

Abstract: A semantic pattern extraction system can distill tremendous amounts of silicon wafer manufacturing data to generate a small set of simple sentences (semantic patterns) describing physical design geometries that may explain manufacturing defects. The system can analyze many SEM images for manufacturing defects in areas of interest on a wafer. A tagged continuous itemset is generated from the images, with items comprising physical design feature values corresponding to the areas of interest and tagged with the presence or absence of a manufacturing defect. Entropy-based discretization converts the continuous itemset into a discretized one. Frequent set mining identifies a set of candidate semantic patterns from the discretized itemset. Candidate semantic patterns are reduced using reduction techniques and are scored. A ranked list of final semantic patterns is presented to a user. The final semantic patterns can be used to improve a manufacturing process.

Abstract: A method is provided for populating a map with a set of avatars through the use of a mobile technology platform associated with a user. The method (201) includes developing a set of facial characteristics (205), wherein each facial characteristic in the set is associated with one of a plurality of individuals that the user has encountered over a period of time while using the mobile technology platform; recording the locations (207) and times at which each of the plurality of individuals was encountered; forming a first database by associating the recorded times and locations at which each of the plurality of individuals was encountered with the individual's facial characteristics in the set; generating a set of avatars (309) from the set of facial characteristics; and using the first database to populate (319) a map (307) with the set of avatars.

Abstract: Provided herein are methods, systems, and computer program products for calculating a refined segmentation of an image which identifies a foreground of the image and a background of the image. Embodiments include creating a refined segmentation of an image by accessing an image that includes a foreground image and background information; creating a first segmentation of the image by applying an image subtraction algorithm; creating a second segmentation of the image by applying a semantic segmentation algorithm; creating an initialization segmentation by combining the first segmentation and the second segmentation; and creating a resulting refined segmentation by refining the initialization segmentation using at least one of statistical analysis or graphical methods. Statistical analysis includes applying a Gaussian mixture model and graphical methods include a graph cut global optimization.

Abstract: Systems and methods are described to enable the creation of user interfaces that may adapt to different environments and may be automatically created. User interfaces may be two-dimensional or three-dimensional and may be used in virtual reality or augmented reality applications. An interface creator may create or receive digital assets associated with a content item, define virtual planes and associated digital asset templates, associate the digital assets with the virtual planes, and enable display of the virtual planes with associated digital assets to a user for user interaction. Digital assets may be automatically edited to meet the specifications of the templates associated with the virtual planes. Virtual planes and templates may also be standardized and aggregated so that a completed user interface may be easily delivered and presented with other content items in a uniform manner.

Abstract: A radiographic image capturing system is described. If a hardware processor determines that an image processor and an image analyzing unit are capable of sharing image data via an identical memory, the image processor stores image data in the memory and the image analyzing unit analyzes the image data with reference to the memory. If the hardware processor determines that the image processor and the image analyzing unit can send and receive the data via a wired network, the image processor transfers the data to the image analyzing unit, and the image analyzing unit analyzes the data. If the hardware processor determines that the image processor and the image analyzing unit can send and receive the data via a wireless network, the image processor compresses the data or decimates partial data and transfers the data to the image analyzing unit, and the image analyzing unit decompresses and analyzes the data.

Abstract: A data generation apparatus includes: an acquisition unit configured to acquire a result determined by a determination unit, which uses an identification device trained using learning data, as to whether or not an object to be inspected includes a part to be detected based on an image of the object to be inspected; an evaluation unit configured to evaluate whether or not the determination result is correct; and a generation unit configured to generate, if the evaluation unit has evaluated that the determination result is not correct, new learning data by associating at least one of the image and a composite image generated based on the image with information in which the determination result is corrected.

Abstract: A machine is configured to determine fashion preferences of users and to provide item recommendations based on the fashion preferences. For example, the machine accesses an indication of a fashion style of a user. The fashion style is determined based on automatically captured data pertaining to the user. The machine identifies, based on the fashion style, one or more fashion items from an inventory of fashion items. The machine generates one or more selectable user interface elements for inclusion in a user interface. The one or more user interface elements correspond to the one or more fashion items. The machine causes generation and display of the user interface that includes the one or more selectable user interface elements. A selection of a selectable user interface element results in display of a combination of an image of a particular fashion item and an image of an item worn by the user.

Abstract: An association method executed by a computer, the association method including obtaining a plurality of images captured by a plurality of cameras, acquiring at least one identifier of at least one radio terminal that is located near one of the plurality of cameras when at least one of the plurality of cameras captures one of the plurality of images, obtaining attributes corresponding to one or more person included in a group of images based on an analysis regarding the group of images, the group of images being included in the plurality of images and corresponding to a first identifier included in the at least one identifier, determining an attribute corresponding to the first identifier based on a result of the obtaining the attributes, and storing, into a storage device, the determined attribute in association with the first identifier.

Abstract: The present disclosure provides a method and a device for generating a briefing, the method includes: determining a target news category of a user according to historical accessing data of the user; matching the target news category with a preset event map, to obtain a plurality of pieces of news related to the target news category; extracting a news abstract from each news, to obtain a plurality of news abstracts; selecting a part of news abstracts from the plurality of news abstracts, and generating the briefing based on the part of news abstracts.

Abstract: Provided herein are systems, methods and computer readable media for classification and tagging of textual data. An example method may include accessing a corpus comprising a plurality of documents, each document having one or more labels indicative of services offered by a merchant, generating a query based on extracted features and the documents, generating a precision score for at least a portion of the generated query and selecting a subset of the generated queries based on an assigned precision score satisfying a precision score threshold, the selected subset of the generated queries configured to provide an indication of one or more labels to be applied to machine readable text. A second example method, utilized for tagging machine readable text with unknown labels, may include assigning a label to textual portions of the machine readable text based on results of the application of the queries.

Abstract: A method of automatically detecting anomaly from aerial images of an object of interest is provided. The method may include generating a data coding model corresponding to a category of assets by training a neural network with a training set of digital images depicting an asset in a state that is free from anomalies. The method may further include receiving a target digital image depicting a target asset, and reconstructing the target digital image using the data coding model to generate a decoded target digital image associated with the state that is free from anomalies. The data coding model may be self-supervised to learn to reconstruct itself to an anomaly-free state. The method may also include comparing the target digital image to the decoded target digital image to generate a difference map and, in response to a determination that the difference map depicts any anomaly, generating anomaly alert data.

Abstract: An information providing apparatus (100) includes: an organism recognizing unit (12) recognizing an organism in an outside world; a harmfulness level detecting unit (14) detecting a harmfulness level of the organism recognized by the organism recognizing unit (12); an emotion generating unit (15) generating an emotion that a device (1) is assumed to have in accordance with the harmfulness level detected by the harmfulness level detecting unit (14); and an emotion output unit (16) outputting the emotion produced by the emotion generating unit (15) in a mode recognizable by a user.

Abstract: Methods and systems for a circuit similarity metric for semiconductor testsite coverage. One or more unique values for each of a set of measures for each circuit layout of a plurality of circuit layouts are identified and a pairwise comparison across the set of measures is conducted for a selected pair of the plurality of circuit layouts to derive a similarity score for the selected pair of circuit layouts. The similarity score is incremented for the selected pair in response to the selected pair of circuit layouts sharing a same unique value and the similarity score is decremented for the selected pair in response to one circuit layout of the selected pair of circuit layouts having a unique value that the other circuit layout of the selected pair does not contain.

Abstract: An object recognition ingestion system is presented. The object ingestion system captures image data of objects, possibly in an uncontrolled setting. The image data is analyzed to determine if one or more a priori know canonical shape objects match the object represented in the image data. The canonical shape object also includes one or more reference PoVs indicating perspectives from which to analyze objects having the corresponding shape. An object ingestion engine combines the canonical shape object along with the image data to create a model of the object. The engine generates a desirable set of model PoVs from the reference PoVs, and then generates recognition descriptors from each of the model PoVs. The descriptors, image data, model PoVs, or other contextually relevant information are combined into key frame bundles having sufficient information to allow other computing devices to recognize the object at a later time.

Abstract: A device, system and method for determining incident objects in secondary video is provided. An electronic device: plays, at a display screen, a primary video from a primary video source located at a primary location; determines an incident type associated with the primary video; determines incident object(s) associated with the incident type; detects a particular incident object in a secondary video from a secondary video source located at a secondary location within a threshold distance of the primary location, the particular incident object appearing in the secondary video within a threshold time period from a current playback time of the primary video; determines a relative direction of a location of the particular incident object from the primary location; and controls the display screen to render a particular visual indicator of the particular incident object over or adjacent the primary video at a position indicative of the relative direction.

Abstract: Systems, methods, and non-transitory computer-readable media can train a machine learning model to output respective scores for content items based at least in part on information describing a user, wherein a score for a content item measures a likelihood that the user will select the content item to be included in a social profile of the user. A determination is made that a first user of the social networking system is eligible for a content item suggestion. A first content item to be provided as a suggestion to the first user is determined based at least in part on the model. The first content item is provided as a suggestion to the first user for use in a social profile of the first user.

Abstract: Described herein are systems, methods, and apparatus for automatically identifying and recovering individual cells of interest from a sample of biological matter, e.g., a biological fluid. Also described are methods of enriching a cell type of interest. These systems, methods, and apparatus allow for coordinated performance of two or more of the following, e.g., all with the same device, thereby enabling high throughput: cell enrichment, cell identification, and individual cell recovery for further analysis (e.g., sequencing) of individual recovered cells.

Abstract: An information processing apparatus includes an acquisition unit to acquire layout data in which images are arranged on each page of a photo book from an image group, a division unit to divide the image group into subsets based on a number of pages of the photo book, and an image exchange unit to exchange an image arranged within the layout data. The image exchange unit exchanges a first image arranged within a page corresponding to a first subset and in which the first person is included, with a second image in which a second person, different from the first is included, when the second image is included within the first subset, and exchanges the first image with a third when no person is included, when the second image is not included within the first subset but the third image is included.

Abstract: In an illustrative embodiment, systems and methods for generating data metrics and relationship analysis from an organization's electronic communications archives include pre-processing the electronic communications into a consistent, workable format, including filtering the data to remove irrelevant messages. Machine learning models may be applied to support automatic identification of relevant message content for data analytics. The systems and methods may link the electronic communications with transaction records of a transactional platform and analyze the communications traffic in view of transactional patterns and outcomes. Communications between parties may be analyzed to identify timings and patterns, plus correlations between electronic communication patterns and business outcomes.

Abstract: A resolving method and a system, the method includes: creating a sample library by utilization of an original high-resolution (HR) image set; training a convolutional structural network by utilization of the sample library; and obtaining an HR output signal by processing a low-resolution (LR) input signal by utilization of the trained convolutional structural network. In the resolving method and system according to this disclosure, data after expansion may be processed by simple expansion hardware, without a large change algorithm; and complex algorithms are allocated into parallelizing design, and different servers operate mutually independent; and also, due to the modular design, the design proposals of functional modules may be modified by latter optimization.

Abstract: A system includes determination of a first interval encoding of a first version of a hierarchy of nodes, determination of a second version of the hierarchy of nodes, determination of a first difference between the first version of the hierarchy of nodes and the second version of the hierarchy of nodes, and determination of a first swap of two translation ranges of the first interval encoding, the first swap representing the first difference.

Abstract: A clock fail detector is provided. The clock fail detector includes a timing control signal generator and a clock fail detection module, which may generate control signals according to a clock signal and perform clock fail detection according to the control signals, respectively. The clock fail detection module may comprise first integrators, sample and hold circuits, a second integrator and a comparator. The first integrator may convert previous periods of the clock signal into reference voltages according to ping pong mode control signals within the control signals, respectively. The sample and hold circuits may sample and hold the reference voltages according to the ping pong mode control signals. The second integrator may convert a current clock period of the clock signal into a ramp signal. The comparator may compare the ramp signal with a reference voltage to generate a comparison result signal for indicating whether the clock signal is normal.

Abstract: Embodiments of the present disclosure disclose an artificial intelligence based method and apparatus for classifying a voice-recognized text. A specific embodiment of the method includes: acquiring a current interactive text of a voice query from a user; analyzing the current interactive text using a lexical analyzer to obtain a current lexical structure; determining whether the current lexical structure matches a template of a category in a classifier; and classifying, if the current lexical structure matches the template of the category in the classifier, the current interactive text corresponding to the current lexical structure into the category belonging to the matched template. The embodiment can fast classify texts, effectively reduce the magnitude of manually annotated texts, and improve the annotation efficiency in intelligent voice interaction services.

Abstract: A virtual viewpoint image generation apparatus including: a first generation unit configured to generate, based on a plurality of captured images obtained by a plurality of first cameras, a first virtual viewpoint image in accordance with a position and direction of a virtual viewpoint; a determination unit configured to determine, in accordance with evaluation results of the first virtual viewpoint image, whether or not to generate a second virtual viewpoint image whose image quality is higher than that of the first virtual viewpoint image based on one or a plurality of captured images obtained by one or a plurality of second cameras; and a second generation unit configured to generate the second virtual viewpoint image whose image quality is higher than that of the first virtual viewpoint image in accordance with determination by the determination unit.

Abstract: An identification method includes acquiring image data, acquiring a photography condition for the image data, selecting a classifier corresponding to the acquired photography condition from a plurality of classifiers, each of which was created for one of a plurality of different photography conditions, and identifying a target in the acquired image data by using the selected classifier.

Abstract: An interactive data visualization system is provided that utilizes unsupervised learning process, to automatically choose the hyperparameters for generating insights, which are then used for visualizing the data using interactive plots that update dynamically in response to input control commands.

Abstract: A portable multifunction device displays a first user interface object and a second user interface object on a touch screen display. Upon detecting a finger-down event at the first user interface object and one or more finger-dragging events on the touch screen display, the device moves the first user interface object on the touch screen display along a path determined by the finger-dragging events until the first user interface object at least in part overlaps the second user interface object. Upon detecting a finger-up event at the second user interface object, the device visually replaces the second user interface object with the first user interface object.

Abstract: A method for implementing Live Documentation (LiDo) systems to incorporate and authenticate metadata of electronic documents. An electronic document received at a first LiDo system is partitioned into textual units. Each textual unit has a payload with a printable message and metadata that is not visible to the consumer, wherein the metadata includes one or more attributes. A second LiDo system uses the one or more attributes of a textual unit to authenticate the writer of the textual unit and filter textual units to display only printable messages with a validity above a designated threshold.

Abstract: A method of detecting a back of a shelf for supporting objects includes: obtaining an image depicting a shelf having a shelf edge and a support surface extending from the shelf edge to a shelf back; decomposing the image into a plurality of patches; for each patch: generating a feature descriptor; based on the feature descriptor, assigning one of a shelf back classification and a non-shelf back classification to the patch; generating a mask corresponding to the image, the mask containing an indication of the classification assigned to each of the patches; and presenting the mask.

Abstract: A plurality of binary support vector machines (SVM) are trained to vote between two of a plurality of classes. Each of the binary SVMs vote between a different pair of the plurality of classes. A set of validated samples is input to each of the binary SVMs. The set of validated samples includes samples of each of the classes. Likelihood estimates are determined based on votes of the binary SVMs. A prior probability distribution (prior) is assigned for each of the classes. A posterior probability is calculated for each of the classes based on the determined likelihood estimates and the assigned priors.

Abstract: [Object] To easily search a plurality of captured images for a captured image in which a subject to be imaged that has once appeared in a captured image appears again is desired. [Solution] An information processing apparatus according to the present disclosure includes: a reading unit configured to read, from a recording unit configured to record radio wave information concerning a radio wave, a captured image, and a time at which the captured image is taken in correlation with one another, the radio wave information; and a searching unit configured to search a plurality of the captured images recorded in the recording unit for a captured image that satisfies a search condition on a basis of at least one of a number of radio waves included in the read radio wave information and an intensity of the radio waves.

Abstract: A method of recognizing edges in a camera image (A) produced by a camera, includes steps of: assigning (S1) a respective scaling size to each pixel (aij) of the camera image (A) as a function of an object distance, relative to the camera, of a spatial region imaged in the respective pixel (aij); producing (S2) an integral image (B) from the camera image (A); applying (S3) Haar wavelet transforms to the pixels (bij) of the integral image (B) to generate a gradient image (C), wherein a scaling of the Haar wavelets is determined for each pixel (bij) of the integral image (B) taking account of the scaling size assigned to the corresponding pixel (aij) of the camera image (A); and recognizing (S4) edges in the camera image (A) using the generated gradient image (C).

Abstract: According to one implementation, an image processing system includes a computing platform having a hardware processor and a system memory storing a software code including a convolutional neural network (CNN) trained using one or more semantic map(s). The hardware processor executes the software code to receive an original image including multiple object images each identified with one of multiple object classes, and to generate replications of the original image, each replication corresponding respectively to one of the object classes. The hardware processor further executes the software code to, for each replication, selectively modify one or more object image(s) identified with the object class corresponding to the replication, using the CNN, to produce partially modified images each corresponding respectively to an object class, and to merge the partially modified images, using the CNN, to generate a modified image corresponding to the original image.

Abstract: The present invention concerns a method for assigning a pixel to one of a foreground pixel set and a background pixel set using asynchronous semantic post processing to improve motion detection by an imaging device, the pixel belonging to an image of a chronological sequence of images taken by the imaging device that includes background and foreground objects.

Abstract: The invention provides a saturation enhancement method and device based on fuzzy control. The saturation control method based on fuzzy control utilizes a fuzzy system to output an enhancement weight according to the saturation component, value component and detail information or saturation component, intensity component, and detail information of an input image, and feeds back the enhancement weight to the enhancement function, to change the enhancement intensity of the saturation component. As such, the color uncertainty caused by saturation enhancement can be avoided, and the loss of detail information in saturation enhancement can be reduced.

Abstract: Systems and methods are provided for automatic classification of multiple abnormalities that are visible in chest X-ray images. The systems and methods are based on a deep learning architecture that predicts, in addition to classification scores of abnormalities, lung/heart masks, and the location of certain abnormalities. By training a multi-task network to improve all the results, the network and the resulting abnormality classification is improved. Normalization of the chest X-ray images is also used to improve the accuracy and efficiency of the multi-task network.

Abstract: Photographs can sometimes be altered or changed in order to create photographs which appear to be of a scene which never took place, or did not take place at the place and time when the photograph is purported to have been taken. One way of circumventing known systems to detect this sort of falsified image is to take a photograph of a printed altered image. The present invention is a method of detecting such “photographs of photographs” by comparing two photographs of the same seen taken at the same time, with different exposures.

Abstract: The invention relates to an image processing device (10) for processing diagnostic image data, comprising: a data input (11) for receiving vector-valued diagnostic image data and a quantification unit (12) for, for each pixel, determining a subset of identifiers, selected from a predetermined set of template identifiers, and determining for each identifier of the subset a quantification value indicative of a presence, proportion or significance in the pixel of a material or condition corresponding to the identifier. The device comprises a dictionary definition unit (14) for providing a dictionary that assigns an index to each unique identifier subset, and an image data encoder (16) for encoding the image data. The encoder is adapted for, for each pixel, calculating a bit sequence comprising a first set of bits encoding the index and further sets of bits encoding quantification values.

Abstract: A surveillance system may comprise one or more computing devices and one or more robotic surveillance devices. The one or more computing devices may be configured to obtain video data captured by one or more cameras. The one or more computing devices may analyze the video data to determine whether there is any trigger event. In response to determining that there is a trigger event, the one or more computing device may determine an optimal robotic surveillance device among the one or more robotic surveillance devices based on the trigger event and provide an instruction to the optimal robotic surveillance device. The optimal robotic surveillance device may be configured to perform a responding action in response to receiving the instruction.

Abstract: Methods and systems disclosed herein may quantify a representation of a type of input an image analysis system should expect. The image analysis system may be trained on the type of input the image analysis system should expect using a first image stream. A first model of the type of input that the image analysis system should expect may be built from the first image stream. After the first model is built, a second image, or a second image stream, may be compared to the first model to determine a difference between the second image, or second image stream, and the first image stream. When the difference is greater than or equal to a threshold, a drift may be detected and steps may be taken to determine the cause of the drift.

Abstract: The invention includes: a characteristic extraction unit for extracting facial characteristics of persons from moving image data; correspondence information extraction unit for extracting characteristic information corresponding to facial characteristics extracted by characteristic extraction unit from pre-stored characteristic information on faces of multiple persons; exclusion candidate setting unit for setting, to an exclusion candidate, characteristic information selected in accordance with an input of a user or a predetermined criterion among analogous characteristic information if multiple characteristic information extracted by the correspondence information extraction unit is analogous; and a person information output unit for displaying, on a display unit, information indicating a person corresponding to characteristic information extracted by the correspondence information extraction unit and set to the exclusion candidate by the exclusion candidate setting unit, as well as information indicating a

Abstract: The present disclosure relates to apparatus and method for training a learning system to detect event. The present disclosure provides apparatus and method for training a learning system that includes an event related keyword collecting unit that collects an event related keyword inputted by a user, a related keyword collecting unit that collects at least one related keyword from a word database by transmitting an event related keyword to the word database, an event related image collecting unit that collects at least one event related image that is related to a search formula from an image database by transmitting the search formula to the image database, and a training unit that trains a learning system by communicating with the learning system with a use of an event related image as training data.

Abstract: Fast classification of data can be performed according to characteristics. In a clustering device (100), a data storage unit (300) stores a plurality of data sets. A cluster generation unit (400) generates an approximate line that approximates as many data sets as possible within a predetermined margin of error among the plurality of data sets in a space in which the plurality of data sets are arranged in accordance with data values. The cluster generation unit (400) generates a cluster by classifying the plurality of data sets based on the generated approximate line and outputs the generated cluster.

Abstract: A computer-implemented method, including receiving, by a monitoring system that is configured to monitor a property and from a sensor of the monitoring system, sensor data, receiving video data, determining that a vehicle is in a vicinity of the property based on the sensor data and the video data, comparing frames of the video data to an image of a known vehicle, determining a confidence score that reflects a likelihood that the vehicle in the vicinity of the property matches the known vehicle, based on determining that the confidence score that reflects the likelihood that the vehicle in the vicinity of the property matches the known vehicle satisfies a confidence score threshold, identifying an action that is performed by the vehicle, identifying a monitoring system action that corresponds to the action performed by the vehicle, and performing the monitoring system action.

Abstract: A learning procedure involves generating a non-defective product learning model by conducting machine learning using non-defective product data as teacher data, and generating a defective product learning model for each defect type by conducting machine learning for each defect type using defective product data as teacher data. A calculating procedure involves calculating the likelihood of a non-defective product from output data calculated using the non-defective product learning model to which target product data is input, and calculating the likelihood of a defective product for each defect type from output data calculated using the defective product learning model to which the target product data is input. A determining procedure involves determining that the target product data is data on a defective product having an unknown defect when the likelihood of a non-defective product and the likelihood of a defective product for each defect type satisfy a predetermined requirement.