Abstract: Techniques are generally described for performing outfit recommendation using a recurrent neural network. In various examples, a computing device may receive a first state vector representing an outfit comprising at least one fashion item. First image data depicting a second fashion item of a first item category may be received. A machine learning mode may generate a first output feature vector based on the first state vector, the first image data, a first attribute vector, and the first item category. The first output feature vector may be compared to other feature vectors representing other fashion items in the first category to determine distances between the first output feature vector and the other feature vectors. A set of fashion items may be recommended based on the distances between the first output feature vector and the other feature vectors.

Abstract: A monocular snapshot four-dimensional imaging method and system are provided. The method includes constructing a numerical simulation model of an imaging system, and optimizing a phase distribution of an entrance pupil plane of the imaging system to determine a point spread function and a lens parameter corresponding to the phase distribution; building the imaging system, calibrating the point spread function, and obtaining a distortion parameter of the imaging system; taking a single shot on a target scene to obtain a single encoded image containing four-dimensional light field information of the target scene, and correcting the single encoded image according to the distortion parameter of the imaging system; and decoding the four-dimensional light field information of the target scene from the single encoded image according to the calibrated point spread function to obtain an intensity image, a polarization contrast image and a depth map of the target scene.

Abstract: A method comprising: receiving a dataset comprising a plurality of data instances; extracting a feature vector representation of each of the data instances in the dataset; choosing a first data instance for adding to a subset of the dataset, wherein the first data instance is removed from the dataset; performing an iterative process comprising: (i) identifying one of the data instances in the dataset which represents a maximal information addition to the subset, based, at least in part, on measuring an information difference parameter between the feature vector representation of the identified data instance and the feature vector representations of all of the data instances in the subset, and (ii) adding the identified data instance to the subset and removing the identified data instance from the dataset, until the information difference parameter is lower than a predetermined threshold; and outputting the subset as a representative subset of the dataset.

Abstract: A method for detecting defects in product applied in a computer device inputs an image of a product under test to an automatic encoder to obtain a reconstructed image, and the image is segmented into N image blocks and the reconstructed image is segmented into N image blocks. The computer device associates each of the N testing blocks with one reconstructed blocks according to positions of the N testing blocks in the image and positions of the N reconstructed blocks in the reconstructed image. The computer device further calculates mean square errors between each of the N testing blocks and each of the N reconstructed blocks, and associates each mean square error with each of the N testing blocks, whether the product has defects being determined based on the mean square errors corresponding to each of the N testing blocks.

Abstract: Provided are methods for automatically detecting traffic signals using sensor data. The methods can include obtaining first sensor data and second sensor data regarding an environment of an autonomous vehicle, where the first sensor data represents at least one image of the environment generated by at least one image sensor of the autonomous vehicle, where the second sensor data represents a three-dimensional point cloud associated with the environment, and where the second sensor data is generated by at least one range sensor of the autonomous vehicle. The method can also include determining a portion of the at least one image corresponding to a traffic signal in the environment, determining a point cluster of the point cloud corresponding to the portion of the at least one image, and determining a location of the traffic signal based on the point cluster. Systems and computer program products are also provided.

Abstract: Systems and methods monitor a workspace for safety purposes using sensors distributed about the workspace. The sensors are registered with respect to each other, and this registration is monitored over time. Occluded space as well as occupied space is identified, and this mapping is frequently updated. Based on the mapping, a constrained motion plan of machinery can be generated to ensure safety.

Abstract: A trained artificial intelligence engine analyzes a subject's dental images and dental survey results to determine whether the subject has an apparent dental condition. The artificial intelligence engine is trained using manually-classified dental images and model survey results. The trained artificial intelligence engine can annotate the subject's dental images to indicate landmark structures and the location of the apparent dental condition. In addition, the trained artificial intelligence engine can generate a confidence score that indicates the likelihood that the subject has the apparent dental condition.

Abstract: An apparatus includes a memory and processor. The memory stores document categories, text generated from an image a physical document page, and a machine learning algorithm. The machine learning algorithm is configured to extract features associated with natural language processing and features associated with the text. The machine learning algorithm is also configured to generate a feature vector that includes the first and second pluralities of features, and to generate, based on the feature vector, a set of probabilities, each of which is associated with a document category and indicates a probability that the physical document from which the text was generated belongs to that document category. The processor applies the machine learning algorithm to the text, to generate the set of probabilities, identifies a largest probability, and assigns the image to the associated document category.

Abstract: A method is disclosed for determining a position of a first image region in an image corresponding to a second image region in the image. The method comprises: providing the first image region comprising a reference position; providing the image; determining a plurality of signature values for different image positions in the first image region; assigning relative offsets of a plurality of image positions in the first image region to respectively determined signature values; determining a plurality of correspondence signature values to different image positions in the image; determining relative offsets relative to determined correspondence signature values with use of the assignment; calculating target positions by summing the relative offset and corresponding image position; incrementally assigning a weighting value to a calculated target position, thus forming an incremental weighting value; and determining the position of the first image region by finding an extremum in the incremental weighting values.

Abstract: An iris recognition apparatus for performing recognition of a target person based on an iris pattern of the target person, is provided with: an imaging device that performs an imaging processing for sequentially outputting imaged images while changing a focal position; and a determining device that starts a focusing determination for determining whether or not an image is a focused image before the imaging device completes the imaging processing.

Abstract: The present disclosure is directed to devices and methods for synchronizing capturing of spectral images with the capturing of thermal images. A thermal imaging device of an aerial vehicle captures a sequence of thermal image of thermal images. Capturing of spectral images by a spectral imaging device of the aerial vehicle is synchronized with the capturing of the thermal images. Irradiance data indicative of a background temperature is sensed. A digital surface model of an area of interest is generated based on the sequence of spectral images. An emissivity of a target is estimated and a temperature of a pixel of the digital surface model of the target is estimated based on the sequence of thermal images, the irradiance data indicative of the background temperature and the estimated emissivity of the target.

Abstract: A connected layer feature is generated by connecting outputs of a plurality of layers of a hierarchical neural network obtained by processing an input image using the hierarchical neural network. An attribute score map representing an attribute of each region of the input image is generated for each attribute using the connected layer feature. A recognition result for a recognition target is generated and output by integrating the generated attribute score maps for respective attributes.

Abstract: Methods, systems, and computer-readable storage media for determining that a subject is a live person include obtaining, by an image capture device, a set of subject images. Each image is captured at a different corresponding relative location of the image capture device with respect to the subject. Parameters are determined from the set of images of the subject. The parameters represent corneal reflections of at least one object in at least one eye of the subject. A determination is made, based on the parameters, that the subject is a live person. Responsive to determining that the subject is a live person, an authentication process is initiated to authenticate the subject.

Abstract: The majority of image reconstruction algorithms are common to DT and CT, and require reconstruction volume allocation and are based on ray tracing techniques. Reconstructed three-dimensional images become available only after the entire volume is processed and the algorithm completes. The present invention performs reconstruction on a slice-by-slice basis, instead of waiting for completion of the algorithm by back-projecting each pixel in each attenuation image towards the emitter that generated that image, onto a selected reconstruction slice and determining a proportion of overlap with grid cells in the slice to obtain weighting factors in order to calculate an average back-projected intensity for each grid cell in the selected reconstruction slice.

Abstract: A text correction engine meets different and changing end user requirements, with the ability to change a desired output by providing sufficient amounts of data, and by finetuning the appropriate text correction engine at the point of origin of the data. It is possible to retain confidentiality of data by retraining the base deep learning model at the base deep learning model's point of origin, to improve the base deep learning model's performance, making the base deep learning model more accurate for different contexts. Separate training of an end user model, leaving the base deep learning model intact, streamlines end user model training, and highlights desirable changes in the base deep learning model for further training or retraining.

Abstract: A face authentication system includes a registration device that acquires a face image for registration of a user in a facility and specific information used for specifying the user, one or more face authentication machines that acquire a face image for authentication of the user captured when the event is generated, a face management server that is connected to the registration device via a network, and accumulates the face image for registration and the specific information in association with each other, and a face collation server that generates face feature data from the face image for registration and the face image for authentication, and performs collation processing between pieces of the face feature data.

Abstract: The disclosure is inputting a first image captured an image of an authentication target; inputting a second image captured an image of a right eye or a left eye of the target; determining whether the second image is of a left eye or a right eye of the target based on information including the first image, and outputting a determination result as left/right information in association with the second image; detecting an overlap between a region including the second image and a predetermined region in the first image; calculating a verification score by comparing characteristic information that are related to the left/right information with iris characteristic information calculated from the second image, and calculating a first weighted verification score obtained by weighting the verification score with a detection result; and authenticating a target in the second image based on the first weighted verification score, and outputting an authentication result.

Abstract: A device for characterizing a fire comprises at least one stereovision system and at least one processing unit. The at least one stereovision system comprises a first and a second image capture unit. The at least one processing unit is configured to determine at least one geometric characteristic of the fire. The processing unit is also configured to determine a radiative flux of this fire on the basis of a calibrated linear relationship established between a radiative flux of a reference fire as a function of at least one of the geometric characteristics of the fire and of at least one fire category in order to be able to determine zones exposed to a radiative flux that exceeds a reference threshold. The present disclosure also relates to a method for determining radiative fluxes that implements such a fire characterizing device.

Abstract: Disclosed herein are systems and methods for of assessing stain titer levels. An exemplary method includes generating a set of field of views for the image or the region of the image, selecting field of views from the set of field of views that meet predefined criteria, creating a series of patches within each of the selected field of views, retaining patches from the series of patches that meet predefined criteria indicative of a presence of the stain for which the titer is to be estimated, deriving stain color features and stain intensity features pertaining to the stain from the retained patches, estimating a titer score for each of the retained patches based on the stain color features and the stain intensity features, and calculating a weighted average score for the titer of the stain based on the estimated titer score for each of the retained patches.

Abstract: Systems and methods for orthognathic surgical planning are described herein. An example computer-implemented method can include generating a composite three-dimensional (3D) model of a subject's skull, defining a primal reference frame for the composite 3D model, performing a cephalometric analysis on the composite 3D model to quantify at least one geometric property of the subject's skull, performing a virtual osteotomy to separate the composite 3D model into a plurality of segments, performing a surgical simulation using the osteotomized segments, and designing a surgical splint or template for the subject.