Abstract: A method for facial image recognition is provided. A plurality of original facial images are received. A plurality of standard facial images corresponding to the original facial images are generated through a standard face generation model. A recognition model is trained by using the original facial images and the standard facial images. The recognition model is tested by using the original facial image test set and a standard facial image test set until the recognition model recognizes that the first accuracy rate of the original facial image test set is higher than a first threshold value and the second accuracy rate of the standard facial image test set is higher than a second threshold value. The original facial image test set is composed of the original facial images obtained by sampling, and the standard facial image test set is composed of the standard facial images obtained by sampling.

Abstract: A spacer supportability evaluation method and device as well as a computer readable storage medium are provided. The method includes acquiring initial distribution images of spacers and corresponding support pads on a substrate, performing binary grayscaling processing to obtain distribution images of spacers and corresponding support pads, obtaining two binary matrices according to the distribution images, subjecting the two binary matrices to convolution in a spatial domain or to multiplication in a frequency domain to obtain an equivalent support matrix, calculating a number of elements in the equivalent support matrix whose values are a first value to obtain a number of supported pixels. The supportability of spacers is evaluated by acquiring parameters or design drawings of the spacers to calculate suitable size and positional arrangement of each spacer, improving the supportability of spacers and keeps the cell gap of the liquid crystal cell stable and uniform.

Abstract: A device includes at least one processor and memory including instructions that when executed by the at least one processor, cause the at least one processor to scan an input image including pixels that have either a first state or a second state opposite the first state, initiate a first cluster and a first polygon that belongs to the first cluster upon encountering a first pixel of the input image that has the first state, execute a first set of operations to form the first polygon, generate an indication of one or more characteristics of the input image based on at least the first cluster, and output the indication.

Abstract: Methods and systems for image analysis are provided, and in particular for identifying a set of base-calling locations in a flow cell for DNA sequencing. These include capturing flow cell images after each sequencing step performed on the flow cell, and identifying candidate cluster centers in at least one of the flow cell images. Intensities are determined for each candidate cluster center in a set of flow cell images. Purities are determined for each candidate cluster center based on the intensities. Each candidate cluster center with a purity greater than the purity of the surrounding candidate cluster centers within a distance threshold is added to a template set of base-calling locations.

Abstract: Systems and methods are provided for processing a set of multiple serially acquired magnetic resonance spectroscopy (MRS) free induction decay (FID) frames from a multi-frame MRS acquisition series from a region of interest (ROI) in a subject, and for providing a post-processed MRS spectrum. Processing parameters are dynamically varied while measuring results to determine the optimal post-processed results. Spectral regions opposite water from chemical regions of interest are evaluated and used in at least one processing operation. Frequency shift error is estimated via spectral correlation between free induction decay (FID) frames and a reference spectrum. Multiple groups of FID frames within the acquired set are identified to different phases corresponding with a phase step cycle of the acquisition. Baseline correction is also performed via rank order filter (ROF) estimate and a polynomial fit.

Abstract: A sequence of video frames of an area of interest is obtained. A first background model of the area of interest is constructed based on a first parameter. A second background model of the area of interest is constructed based on a second parameter, the second parameter being different from the first parameter. A difference between the first and second background models is determined. A stationary target is determined based on the determined difference. An alert concerning the stationary target is generated.

Abstract: The present invention concerns a method for an artificial neural network to confirm the recognition of handwritten characters produced by a user. The method comprising: training the artificial neural network with a training data set comprising a first set of characters; collecting handwritten characters of a second set of characters produced by the user; and analysing the collected characters of the second set of characters by using the artificial neural network to obtain a first set of probability values comprising character specific probability values for the collected characters, each character specific probability value indicating the probability that the collected character has been correctly interpreted. The analysis considers at least the manner how the collected characters are handwritten and the appearance of the collected characters to obtain the character specific probability values. The invention also relates to a corresponding data processing apparatus.

Abstract: Systems and techniques that facilitate few-shot temporal action localization based on graph convolutional networks are provided. In one or more embodiments, a graph component can generate a graph that models a support set of temporal action classifications. Nodes of the graph can correspond to respective temporal action classifications in the support set. Edges of the graph can correspond to similarities between the respective temporal action classifications. In various embodiments, a convolution component can perform a convolution on the graph, such that the nodes of the graph output respective matching scores indicating levels of match between the respective temporal action classifications and an action to be classified. In various embodiments, an instantiation component can input into the nodes respective input vectors based on a proposed feature vector representing the action to be classified.

Abstract: A start switch device includes a start button configured to send a command to start or stop a vehicle drive device, a biometric sensor arranged on the start button for reading a biometric information of an operator to operate an operating surface of the start button, a light source configured to emit an illuminating light from an inside of the start button toward the operating surface, and a design part arranged around the biometric sensor on the operating surface and configured to define a design on the operating surface by the illuminating light transmitted therethrough.

Abstract: Aspects of the invention provide a vision pipeline-based method of censusing a crowd that includes presenting content on an outdoor digital display or other content player (e.g., a loudspeaker) and capturing with a video camera or other image acquisition device frames or other time-wise succession of images of a scene in the field of view or otherwise in the vicinity of the display/player. First and second people detectors (both, for example, face detectors) are used to determine respective counts of persons in the scene. Estimated viewing statistics, generated with a detection pipeline that includes the first people detector and a tracker, include at least one of (i) a number of persons in the scene that viewed the content on the player, and (ii) for at least one of those persons, a duration during which he/she was in the scene and/or looking toward the player.

Abstract: An image processing method includes: detecting a plurality of feature lines from a first image captured from a first position; specifying, based on a positional relationship between the plurality of feature lines and a plurality of projection lines generated by projecting each of a plurality of line segments onto the first image, a feature line representing a defective portion of a shape of an object; setting a plurality of candidate positions based on the first position, each of the plurality of candidate positions being a candidate for a second position at which a second image is captured; calculating an evaluation value of each of the plurality of candidate positions; determining any of the plurality of candidate positions based on the evaluation value of each of the plurality of candidate positions; and outputting first information to be used in recognition of the determined candidate position.

Abstract: A processor detects feature lines from an image of an object and generates projection lines by projecting line segments in shape information of the object on the image. The processor generates combinations each including a feature line and a projection line and generates plural sets each including a specific number of combinations. The processor uses each set to estimate a position and an orientation of an imaging device that has captured the image, and evaluates a result of the estimation for each set. When there is a change in a relative position or orientation between the object on the image and a shape represented by the shape information, the processor repeats generation of the plural sets, estimation of the position and the orientation of the imaging device and evaluation of the estimation result. The processor determines the position and the orientation of the imaging device based on the repeated evaluation.

Abstract: The method comprises the steps: of applying the identifier (100?) on an object; of making a first digital image and, subsequently, a second digital image of the identifier; of deriving first and second digital keys from the first and second images, respectively; and of comparing the second digital key with the first digital key; wherein the identifier (100?) is made up of same-diameter microbeads (120) distributed in a random manner on the surface of the object, the microbeads being embedded in a layer (110?) of binder, and deriving the first digital key comprises the steps of applying elliptical regression to a distribution of points representative of the positions of the microbeads in the first image in order to define an ellipse, and of applying a change of reference frame consisting in expressing the positions of said points in a reference frame based on the ellipse.

Abstract: A living body detection device (1) includes: an image acquisition unit (171), a determination unit (173) and a detection unit (174). The image acquisition unit (171) acquires a first image in which a subject irradiated by light in a first wavelength range is imaged, and a second image in which the subject irradiated by light in a second wavelength range is imaged, the second wavelength range being different from the first wavelength range. The determination unit (173) determines whether a relation expressed by luminance of the subject imaged in the first image and luminance of the subject imaged in the second image is a relation exhibited by a living body. The detection unit (174) detects that the subject is a living body in a case where the determination unit (173) has determined that it is the relation exhibited by the living body.

Abstract: Various types of objects or occurrences can be automatically detected in input media being processed using a transcoder. The media content can be analyzed to determine various transitions, such as scene changes, which provide insight into useful locations for performing object recognition. Representative frames subsequent a transition are analyzed to determine whether they are appropriate for image analysis, using factors such as amount of motion, brightness, color, or pixel disparity within the frame. If a representative frame meets the various criteria, that frame is sent to an object recognition service for analysis. The output of the service can be a set of object tags that provide information identifying the object and its location in the media. The output tags can be encoded into the output video or stored to an associated metadata file, among other such options.

Abstract: In a vehicular camera, an image generation unit generates at least one image of a directional view from a vehicle, and an object recognition unit recognizes a target object in the at least one image. A region identification unit identifies, in the at least one image, a specific region in the target object recognized by the object recognition unit. A vehicle control unit performs a vehicle control task based on the recognized target object. A mosaic unit identifies a specific region in the at least one image corresponding to the specific region in the at least one image, and performs a mosaic task to thereby blur a mosaic region in the second image. The mosaic region includes at least the specific region in the at least one image.

Abstract: An illustrative example method of tracking an object includes detecting one or more points on the object over time to obtain a plurality of detections, determining a position of each of the detections, determining a relationship between the determined positions, and determining an estimated heading angle of the object based on the relationship.

Abstract: Example embodiments allow for training of artificial neural networks (ANNs) to generate depth maps based on images. The ANNs are trained based on a plurality of sets of images, where each set of images represents a single scene and the images in such a set of images differ with respect to image aperture and/or focal distance. An untrained ANN generates a depth map based on one or more images in a set of images. This depth map is used to generate, using the image(s) in the set, a predicted image that corresponds, with respect to image aperture and/or focal distance, to one of the images in the set. Differences between the predicted image and the corresponding image are used to update the ANN. ANNs tramed in this manner are especially suited for generating depth maps used to perform simulated image blur on small-aperture images.

Abstract: A method of determining respiratory phase of a living body from a sequence of digitized fluoroscopic images of a living-body region exhibiting respiratory displacement, the method employing programmable computing apparatus and comprising the steps of: (a) in each living-body-region image in the sequence, defining one or more zones with each image having identical image-to-image zone locations, sizes, and shapes; (b) for each image, computing an average pixel intensity for each zone to form a sequence thereof for each zone; (c) for each zone, modifying the average pixel intensities by (i) computing the mean value of the sequence of average pixel intensities for such zone, and (ii) subtracting the mean from each average pixel intensity in the zone; (d) for each zone sequence, computing a figure-of-merit; (e) selecting the zone having the highest figure-of-merit; and (f) using the sequence of pixel intensities of the selected zone to determine respiratory phase.

Abstract: A computer-implemented method for assessing if a character in a sample image is formed from a predefined selection of characters, comprising: processing a sample image with an alignment network to form a corrective transformation; applying the corrective transformation to the sample image to form a transformed image; computing a similarity of the transformed image with a corresponding reference image of a character from a predefined selection of characters to form a similarity score; and declaring the sample image not to comprise the character from the predefined selection of characters if the similarity score is less than a threshold.