Abstract: The present disclosure provides an efficient system for cancelable biometric template protection which is not available in the conventional systems. Initially, the system receives a biometric template from a user. The biometric template is one of, a face image, a fingerprint, an iris image and a palmprint. Further, a vector embedding is computed based on the biometric template using a pretrained neural network. After computing the vector embedding, a key based permutation is computed based on the vector embedding and a permutation key. The permutation key is generated based on an Identification number (ID) associated with the user using a random number based key generation technique. Finally, a protected biometric template is generated by computing a permuted embedding corresponding to the biometric template based on the key based permutation using a plurality of fully connected layers pretrained using a weight based center triplet loss function.

Abstract: The present disclosure relates to an image processing apparatus and an image processing method capable of reducing an increase in a load imposed by processing of multispectral image data. Multispectral image data including components in four or more wavelength bands is simple-coded at a fixed compression rate. Alternatively, coded data generated by simple coding at a fixed compression rate for multispectral image data including components in four or more wavelength bands is simple-decoded. For example, the present disclosure is applicable to an image processing apparatus, an image coding apparatus, an image decoding apparatus, an imaging element, an imaging apparatus, or the like.

Abstract: User interaction concepts, principles and algorithms for gestures involving facial expressions, motion or orientation of body parts, eye gaze, tightening muscles, mental activity, and other user actions are disclosed. User interaction concepts, principles and algorithms for enabling hands-free and voice-free interaction with electronic devices are disclosed. Apparatuses, systems, computer implementable methods, and non-transient computer storage media storing instructions, implementing the disclosed concepts, principles and algorithms are disclosed. Gestures for systems using eye gaze and head tracking that can be used with augmented, mixed or virtual reality, mobile or desktop computing are disclosed. Use of periods of limited activity and consecutive user actions in orthogonal axes is disclosed. Generation of command signals based on start and end triggers is disclosed. Methods for coarse as well as fine modification of objects are disclosed.

Abstract: An eye state detecting method, applied to an electronic apparatus with an image sensor, which comprises: (a) acquiring a detecting image via the image sensor; (b) defining a face range on the detecting image; (c) defining a determining range on the face range; and (d) determining if the determining range comprises an open eye image or a close eye image.

Abstract: A decoder includes a memory and a processor coupled to the memory. The processor is configured to obtain first information from a bitstream, the first information indicating a number of first points in a current picture. The processor is also configured to determine second information provided for each of the first points based on the first information, the second information indicating a difference between a coordinate value of a first point and a coordinate value of a corresponding second point, the corresponding second point being included in a reference picture. The processor is further configured to generate a prediction image of a current block by performing an affine prediction according to the determined second information.

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: Disclosed are a face image and iris image acquisition method and device, a computer-readable readable storage medium and an apparatus. The method includes rotating the first tripod head to force the face lens and the iris lens to be in acquisition positions; capturing a first face image and a first iris image simultaneously by the face lens and the iris lens; and locating the iris in the first iris image, and if no iris is located, determining whether a condition of light-avoiding rotation is satisfied, and if the condition is satisfied, rotating the second tripod head to adjust an angle or a position of the supplementary light source to enable a light spot region to avoid an iris region.

Abstract: Example systems, devices, media, and methods are described for controlling the presentation of one or more virtual or graphical elements on a display in response to bimanual hand gestures detected by an eyewear device that is capturing frames of video data with its camera system. An image processing system detects a first hand and defines an input plane relative to a surface of the detected first hand. The image processing system also detects a series of bimanual hand shapes, including the detected first hand and at least one fingertip of a second hand. In response, the system presents a first movable element on the display at a location that is correlated with the current fingertip location. In addition, the image processing system determines whether the detected series of bimanual hand shapes matches a predefined hand gesture. In response to a matching gesture, the system executes a selecting action of an element nearest the current fingertip location.

Abstract: An information processing device according to the present disclosure includes an inference unit that infers, on the basis of a result of checking a first image against a plurality of images taken in the past, a third image that is an image taken in the past at a position corresponding to a second image to be taken at a next timing of the first image, and a generation unit that generates a fourth image that is an image obtained by correcting the second image on the basis of the third image in a case where the second image is acquired.

Abstract: The technology of this application related to an image processor comprising a plurality of modules, the plurality of modules comprising a first module and a second module, wherein the image processor is configured to receive an input image and output a plurality of mathematical descriptors for characteristic regions of the input image. The first module is configured to implement a first trained artificial intelligence model to detect a set of characteristic regions in the input image; and the second module is configured to implement a second trained artificial intelligence model to determine a mathematical descriptor for each of said set of characteristic regions. The first and second trained artificial intelligence models are collectively trained end to end.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer-storage media, for lighting adaptive navigation. In some implementations, a method includes receiving map data associated with a property; obtaining sensor data; based on the map data and the sensor data, determining a lighting scenario; and based on the lighting scenario, configuring the robotic device.

Abstract: A method of determining an optimal target gemstone to be obtained from a rough gemstone comprises obtaining a first series of 2D images of the rough gemstone; providing a 3D model of a target gemstone to be obtained from the rough gemstone; and generating a second series of 2D images of the target gemstone from the 3D model thereof. The method then comprises comparing the first and second series of 2D images to determine an optimal transformation to be applied to the 3D model of the target gemstone.

Abstract: Embodiments are directed to calibrating multi-view triangulation systems that perceive surfaces and objects based on reflections of one or more scanned laser beams that are continuously sensed by two or more sensors. In addition to sampling and triangulating points from a spline formed by an unbroken line trajectory of a laser beam, the calibration system samples and triangulates a corresponding velocity vector. Iterative reduction is performed on velocity vectors instead of points or splines. The velocity vector includes directions and magnitudes along a trajectory of a scanning laser beam which are used to determine the actual velocities. Translation and rotation vectors are based on the velocity vectors for matching trajectories determined for two or more sensors having offset physical positions, which are used to calibrate sensor offset errors associated with the matching trajectories provided to a modeling engine.

Abstract: A camera apparatus includes control circuitry, where in a training phase, the control circuitry is configured to generate a set of binary images of foliage masks, where each binary image comprises one or more foliage regions demarcated from a non-foliage region. A modified training dataset of color images is generated and a custom neural network model for foliage detection is trained. The control circuitry causes the custom neural network model to not only learn a plurality of features related to foliage but also learn a color variation range of predefined color associated with the plurality of features and utilize combination of the plurality of features related to foliage and the color variation range to obtain trained custom neural network model. The control circuitry is configured to operate the trained custom neural network model to detect one or more foliage regions in a new color image in real time.

Abstract: This disclosure describes a system for automatically updating item image information stored in an item images data store and used for processing captured images to identify items represented in those images. In one implementation, once an identity of an item has been verified, captured images of that item are associated with the item and stored in the item images data store. As a result, the item images data store is updated each time an image of the item is captured and the identity of the item is verified.

Abstract: Disclosed are a bone age estimation method and a bone age estimation apparatus. The bone age estimation method may comprise the steps of: extracting a region of interest including a cervical spine region from a lateral cephalometric radiographic image obtained by imaging a subject's cervical spine, by using a first deep learning model; extracting landmarks from the extracted region of interest by using a second deep learning model; calculating a landmark numerical value on the basis of the extracted landmarks; and providing maturity information of a maturation stage of the cervical spine on the basis of the calculated landmark numerical value.

Abstract: The present disclosure provides a training method and device for an image identifying model, and an image identifying method. The training method comprises: obtaining image samples of a plurality of categories; inputting image samples of each category into a feature extraction layer of the image identifying model to extract a feature vector of each image sample; calculating a statistical characteristic information of an actual distribution function corresponding to each category according to the feature vector of each image sample of the each category; establishing an augmented distribution function corresponding to the each category according to the statistical characteristic information; obtaining augmented sample features of the each category based on the augmented distribution function; and inputting feature vectors of the image samples and the augmented sample features into a classification layer of the image identifying model for supervised learning.

Abstract: A method and system for improving prediction of an incident affecting public safety is described. At a machine learning analytics system, the method includes obtaining a real-time video for a location of a public safety emergency call and predicting a public safety incident therefrom. At a learning processor, the method includes: receiving information related to the public safety emergency call; identifying a public safety incident type based thereon; receiving the prediction of the public safety incident; using a first public safety incident machine learning model based on the identified public safety incident type and comparing the public safety incident prediction with the information related to the public safety emergency call and identifying an incident prediction mismatch therefrom. The learning processor identifies an artifact in the real-time video that is a contributing factor for the incident prediction mismatch; and implements a model improvement action associated with the artifact.

Abstract: Feature descriptor matching is reformulated into a graph-matching problem. Keypoints from a query image and a reference image are initially matched and filtered based on the match. For a given keypoint, a feature graph is constructed based on neighboring keypoints surrounding the given keypoint. The feature graph is compared to a corresponding feature graph of a reference image for the matched keypoint. Relocalization data is obtained based on the comparison.

Abstract: Unsupervised, deep learning of eye-landmarks in a user-specific eyes' image data by capturing an unlabeled image comprising an eye region of a user, using an initial geometrically regularized loss function, training a plurality of convolutional autoencoders on the unlabeled image comprising the eye region of the user to recover a plurality of user-specific eye landmarks, training a convolutional neural network for autoencoded landmarks-based recovery from the unlabeled image, and where the initial geometrically regularized loss function is represented by the formula LAE=?reconLrecon+?concLconc+?sepLsep+?eqvLeqv where LAE is total AutoEncoder Loss, ?reconLrecon is ?-weighted reconstruction loss, ?concLconce is ?-weighted concentration loss, ?sepLsep is ?-weighted separation loss, and ?eqvLeqv is ?-weighted equivalence loss.

Abstract: An electronic device is provided. The electronic device includes at least one processor, and a memory functionally connected to the at least one processor. The memory may store instructions that, when executed, enable the electronic device to obtain a plurality of images, generate a first basic extended image based on first images among the plurality of images, identify at least one first masking area included in the first basic extended image, and generate a first inference image by modifying the at least one first masking area using at least one first inference result, based on the first images and the first basic extended image. An angle of view of the first inference image may be larger than an angle of view of each of the first images.

Abstract: Example posture analysis systems and methods are described. In one implementation, a computing system identifies a deviation of a user's posture from a predetermined posture based on visual data associated with the user. The user is informed of the deviation via a graphical image displayed on a display device. The user is then provided with instructions for correcting the deviation via the graphical image displayed on the display device.

Abstract: Processes and systems for displaying information received in a query response. One system includes a communication device including a user interface and an electronic processor. The electronic processor transmits a query, and receives a response to the query including stored information in each of a plurality of fields of information about an object of interest. The electronic processor receives an image of the object from a camera, identifies visually descriptive text information describing the object, and categorizes at least a portion of the visually descriptive text information into the plurality of fields of information. The electronic processor determines a confidence level for each field of information based on a comparison of the stored information and the visually descriptive text information in each field, and displays the stored information in one or more fields for which the confidence level of the comparison is below a predetermined threshold.

Abstract: A meter readout on a meter has digits including a first digit, a second digit, etc. A sequence of images of the meter is obtained. The images include images of the digits in the meter readout. Automated recognition of the digits in the images result in likelihood arrays indicating the likelihoods for the digit values for the digits imaged in the meter images. Short chains of digit values are identified and spliced together to form a series of single digit, two-digit, three-digits, etc. paths that are built up based on the likelihood arrays. Various criteria are used to discard most of the chains and thereby avoid the combinatorial explosion of possible paths and thereby produce reliable meter readings without consuming considerable computational resources.

Abstract: A video data encoding method for encoding an array of video data values comprises the steps of: frequency-transforming the video data values according to a frequency transform, to generate an array of frequency-transformed values by a matrix-multiplication process using a transform matrix having a data precision of fourteen bits, the frequency transform being a discrete cosine transform; defining a 64×64 transform matrix M64 for a 64×64 DCT transform, the matrix M64 being defined by the accompanying FIGS. 10a-10e and 11; for an N×N transform where N is 2, 4, 8 or 16, subsampling the 64×64 transform matrix M64 to select a subset of N×N values, the subset MN[x][y] of values being defined by: MN[x][y]=M64[x][(2(6-log2(N)))y] for x,y=0 . . . (N?1).

Abstract: The present disclosure provides systems and methods for delivering assistance to impaired users within a virtual conferencing system, and more particularly, to providing assistance to hearing-impaired users to control the virtual conferencing system while using sign language. The systems and methods for providing assistance to impaired users includes initiating a virtual meeting, detecting one or more hand gestures within a predefined area, interpreting the one or more hand gestures to determine one or more commands, and updating one or more controls for the virtual meeting based on the one or more commands. These steps can be implemented to assist hearing-impaired users to manage controls when communicating within a virtual meeting.

Abstract: A heterogeneous image registration method includes: performing edge detection on collected images, in combination with a curvature scale space strategy to extract contour curved segments in an edge image. Implementing a feature point detection strategy based on global and local curvature detecting feature points in the contour curved segments, and obtaining the nearest minimum local curvature of the feature points pointing to starting and end points of the contour, respectively. Calculating the number of neighborhood sampling points and neighborhood auxiliary feature points of neighborhoods on both edges of each of the feature points according to the nearest minimum local curvature. Using neighborhood auxiliary feature points and feature points to form a feature triangle, calculating an angle bisector vector and a main direction corresponding to the feature point in the feature triangle.

Abstract: Systems and methods herein describe a motion retargeting system. The motion retargeting system accesses a plurality of two-dimensional images comprising a person performing a plurality of body poses, extracts a plurality of implicit volumetric representations from the plurality of body poses, generates a three-dimensional warping field, the three-dimensional warping field configured to warp the plurality of implicit volumetric representations from a canonical pose to a target pose, and based on the three-dimensional warping field, generates a two-dimensional image of an artificial person performing the target pose.

Abstract: The present invention relates to a method of fusing at least two different candidate lists into a single candidate list, said at least two different candidate lists being generated by matching biometric data captured from an individual to be identified and biometric data enrolled from candidates to calculate match scores of candidates and ranking said match scores, the method comprising steps of: normalizing the match scores of the candidates of said at least two different candidate lists; ranking at least part of the candidates of said at least two different candidate lists in a single candidate list according to said normalized match scores to fuse said at least two different candidate lists; identifying at least one candidate belonging to multiple lists of said at least two different candidate lists; and raising said identified at least one candidate's rank in said single candidate list.

Abstract: An illustrative image descriptor generation system determines a subset of image descriptors from a plurality of image descriptors that each correspond to a different feature point included within an image. The subset of image descriptors is determined based on geometric proximity, within the image, of respective feature points of the subset of image descriptors to a feature point of a primary image descriptor. The image descriptor generation system then selects a secondary image descriptor from the subset of image descriptors and combines the primary image descriptor and the secondary image descriptor to form a composite image descriptor. Corresponding methods and systems are also disclosed.

Abstract: An endoscope system comprising: a light source device configured to emit illumination light; an electronic endoscope; a processor that includes an image processing unit; and a monitor. The image processing unit includes: a detection unit configured to generate a first blood vessel detection image and a second blood vessel detection image; a separation unit configured to separate the first blood vessel detection image and the second blood vessel detection image into a shallow blood vessel image and a deep blood vessel image; a feature amount calculation unit configured to calculate a shallow blood vessel feature amount regarding the shallow blood vessel image and a deep blood vessel feature amount regarding the deep blood vessel image; and an evaluation unit configured to calculate severity of a lesion of a biological tissue by using the shallow blood vessel feature amount and the deep blood vessel feature amount.

Abstract: Disclosed herein are methods, systems, and computer-readable storage media for generating three-dimensional (3D) images of a scene. According to an aspect, a method includes capturing a real-time image and a first still image of a scene. Further, the method includes displaying the real-time image of the scene on a display. The method also includes determining one or more properties of the captured images. The method also includes calculating an offset in a real-time display of the scene to indicate a target camera positional offset with respect to the first still image. Further, the method includes determining that a capture device is in a position of the target camera positional offset. The method also includes capturing a second still image. Further, the method includes correcting the captured first and second still images. The method also includes generating the three-dimensional image based on the corrected first and second still images.

Abstract: A virtual window configuration method includes the following steps. A processor generates a virtual window. A depth detection sensor generates depth information based on an image. The processor analyzes the depth information to generate a depth matrix. The processor finds a depth configuration block in the image using the depth matrix. A feature point detection sensor generates feature point information for the image. The processor analyzes the feature point information to generate a feature point matrix. The processor finds a feature point configuration block in the image using the feature point matrix. The processor moves the virtual window to the depth configuration block or the feature point configuration block.

Abstract: System and method for detecting the authenticity of products by detecting a unique chaotic signature. Photos of the products are taken at the plant and stored in a database/server. The server processes the images to detect for each authentic product a unique authentic signature which is the result of a manufacturing process, a process of nature etc. To detect whether the product is genuine or not at the store, the user/buyer may take a picture of the product and send it to the server (e.g. using an app installed on a portable device or the like). Upon receipt of the photo, the server may process the receive image in search for a pre-detected and/or pre-stored chaotic signature associated with an authentic product. The server may return a response to the user indicating the result of the search. A feedback mechanism may be included to guide the user to take a picture at a specific location of the product where the chaotic signature may exist.

Abstract: A front-end image acquisition component acquires photographs and/or videos of the exterior of a vehicle traveling on a path between a first predefined location and a second predefined location. In one aspect, an image acquisition system comprises a first camera support structure, a second camera support structure, an image processor, and a memory. Cameras are affixed to each of the camera support structures for acquiring image data of a subject vehicle. Image data is controlled by the image processor and the memory. In another aspect, an image acquisition system comprises a first camera tower, a second camera tower, a camera boom. Cameras are affixed to each of the camera towers.

Abstract: A fingerprint processing device includes a match processing unit configured to determine, based on a first degree, a plurality of feature points having a large value of the first degree among the feature points of a fingerprint specified in a fingerprint image of a matching source, the first degree representing a first distance to other feature points, as representative feature points used in match processing of the fingerprint.

Abstract: In one implementation, a computing device to provide a graphical user interface (GUI) for presenting content includes one or more processors; a network interface to establish a network connection with a remote server system; a display to output a plurality of graphical elements as part of the GUI; one or more input devices to receive user input related to the plurality of graphical elements; and memory storing instructions that, when executed, cause the one or more processors to perform operations including selecting a first content element; selecting second content elements; detecting a third content element comprising a content summary of the second content elements; and outputting, in the GUI on the display, a first graphical element for the first content element and a third graphical element for the third content element, wherein second graphical elements for the second content elements are hidden or collapsed in the GUI.

Abstract: An in-vehicle monitoring device includes a first imaging member, a second imaging member, and a controller. The first imaging member is configured to perform first imaging to capture a first image of one or more occupants in a passenger compartment of the vehicle. The second imaging member is configured to capture a second image of the one or more occupants. The controller is configured to execute a monitoring process on the first image and the second image. The first imaging member and the second imaging member are disposed in a center portion of the vehicle in a vehicle width direction of the vehicle so as to be arranged side-by-side in the vehicle width direction such that at least a portion of an imaging range of the first imaging member and at least a portion of an imaging range of the second imaging member overlap.

Abstract: Example systems, devices, media, and methods are described for controlling the presentation of a series of virtual or graphical elements on a display in response to hand gestures detected by an eyewear device that is capturing frames of video data with its camera system. An image processing system detects a series of hand shapes in the video data and determines whether it matches a predefined hand gesture. Each predefined hand gesture is associated with an action; for example, a leafing gesture is associated with a scrolling action. The system controls the display of a series of virtual elements, in accordance with the associated action. In an example series of hand shapes that includes flexing and extending the fingers of a single hand severally and continually in a leafing motion, the matching predefined leafing gesture is associated with a scrolling action, which displays the series of items in a display order, sequentially and in accordance with the detected speed of the moving fingers.

Abstract: A face authentication apparatus includes a face image acquisition unit that acquires a face image of an authentication target, a collation unit that performs face authentication by calculating similarity between face information of the face image of the authentication target and reference face image of each registered user and comparing the similarity with a threshold, a prediction unit that predicts a change in the similarity on the basis of similarity history on authentication success, and a threshold change unit that changes the threshold on the basis of the prediction result.

Abstract: A method of evaluating a patterning process, the method including: obtaining the result of a first measurement of a first metrology target; obtaining the result of a second measurement of a second metrology target, the second metrology target having a structural difference from the first metrology target that generates a sensitivity difference and/or an offset, of a process parameter of the patterning process between the first and second metrology targets; and determining a value pertaining to the patterning process based on the results of the first and second measurements.

Abstract: Disclosed are methods, apparatus and systems for gesture recognition based on neural network processing. One exemplary method for identifying a gesture communicated by a subject includes receiving a plurality of images associated with the gesture, providing the plurality of images to a first 3-dimensional convolutional neural network (3D CNN) and a second 3D CNN, where the first 3D CNN is operable to produce motion information, where the second 3D CNN is operable to produce pose and color information, and where the first 3D CNN is operable to implement an optical flow algorithm to detect the gesture, fusing the motion information and the pose and color information to produce an identification of the gesture, and determining whether the identification corresponds to a singular gesture across the plurality of images using a recurrent neural network that comprises one or more long short-term memory units.

Abstract: A method includes identifying, within an image that represents a fingerprint, a plurality of keypoints. The method also includes determining a plurality of pixel vectors by transforming, for each respective keypoint, a corresponding two-dimensional (2D) array of pixel values associated with the respective keypoint into a corresponding pixel vector, and obtaining a plurality of filter vectors that represent, for each respective filter of a plurality of filters, a transformation of a corresponding 2D array of filter values of the respective filter into a corresponding filter vector. The method additionally includes convolving the plurality of keypoints with the plurality of filters by multiplying, using a tensor processing hardware, a pixel matrix formed by the plurality of pixel vectors and a filter matrix formed by the plurality of filter vectors, and determining, based on the convolving, a representation of the fingerprint.

Abstract: An eye reconstruction-based eye tracking method and apparatus are provided. The eye tracking method includes generating a reconstructed image by performing eye reconstruction with respect to an input image, determining a difference value between the input image and the reconstructed image, selecting one of the input image, the reconstructed image, and a replacement image as a target image based on the determined difference value, and performing eye tracking based on the target image.

Abstract: A plurality of pixel-based sampling points are identified within an image, wherein sampling points of a pixel are distributed within the pixel. For individual sampling points of individual pixels, a corresponding radiance vector is estimated. A radiance vector includes one or more radiance values characterizing light received at a sampling point. A first machine learning module generates, for each pixel, a corresponding intermediate radiance feature vector, based on the radiance vectors associated with the sampling points within that pixel. A second machine learning module generates, for each pixel, a corresponding final radiance feature vector, based on an intermediate radiance feature vector for that pixel, and one or more other intermediate radiance feature vectors for one or more other pixels neighboring that pixel. One or more kernels are generated, based on the final radiance feature vectors, and applied to corresponding pixels of the image, to generate a lower noise image.

Abstract: The present disclosure describes techniques for creating videos using virtual characters. Creation of a video may be initiated by a user. Camera input comprising a human body of the user may be received. The camera input may be split into a first stream for removing the human body and a second stream for animating a virtual character in the video. An inpainting filter may be applied to remove the human body in real time from the camera input. The inpainting filter may be configured to accelerate texture sampling. Output of the inpainting filter may be blended with images comprised in the camera input to generate camera input backgrounds.

Abstract: User interaction concepts, principles and algorithms for gestures involving facial expressions, motion or orientation of body parts, eye gaze, tightening muscles, mental activity, and other user actions are disclosed. User interaction concepts, principles and algorithms for enabling hands-free and voice-free interaction with electronic devices are disclosed. Apparatuses, systems, computer implementable methods, and non-transient computer storage media storing instructions, implementing the disclosed concepts, principles and algorithms are disclosed. Gestures for systems using eye gaze and head tracking that can be used with augmented, mixed or virtual reality, mobile or desktop computing are disclosed. Use of periods of limited activity and consecutive user actions in orthogonal axes is disclosed. Generation of command signals based on start and end triggers is disclosed. Methods for coarse as well as fine modification of objects are disclosed.

Abstract: Systems, apparatuses, and methods for implementing a safety monitor framework for a safety-critical computer vision (CV) application are disclosed. A system includes a safety-critical CV application, a safety monitor, and a CV accelerator engine. The safety monitor receives an input image, test data, and a CV graph from the safety-critical CV application. The safety monitor generates a modified image by adding additional objects outside of the input image. The safety monitor provides the modified image and CV graph to the CV accelerator which processes the modified image and provides outputs to the safety monitor. The safety monitor determines the likelihood of erroneous processing of the original input image by comparing the outputs for the additional objects with a known good result. The safety monitor complements the overall fault coverage of the CV accelerator engine and covers faults only observable at the level of the CV graph.

Abstract: An exemplary method includes a gesture recognition system determining, based on imagery of a user while the user touchlessly interacts with a user interface of a computer system, a configuration of a plurality of landmarks associated with the user and comparing the configuration of the plurality of landmarks to defined landmark configurations associated with a plurality of defined gestures. Each of the plurality of defined gestures may be associated with a different user input enterable by way of the user interface of the computer system. The method may further include the gesture recognition system selecting, based on the comparing of the configuration of the plurality of landmarks to the defined landmark configurations associated with the plurality of defined gestures, a defined gesture included in the plurality of defined gestures, and directing the computer system to enter a user input that is associated with the defined gesture.

Abstract: A target tracking method is provided for a computer device. The method includes determining a target candidate region of a current image frame; capturing a target candidate image matching the target candidate region from the current image frame; determining a target region of the current image frame according to an image feature of the target candidate image; determining motion prediction data of a next image frame relative to the current image frame by using a motion prediction model and according to the image feature of the target candidate image; and determining a target candidate region of the next image frame according to the target region and the motion prediction data.