Abstract: Described is a method that involves operating an unmanned aerial vehicle (UAV) to begin a flight, where the UAV relies on a navigation system to navigate to a destination. During the flight, the method involves operating a camera to capture images of the UAV's environment, and analyzing the images to detect features in the environment. The method also involves establishing a correlation between features detected in different images, and using location information from the navigation system to localize a feature detected in different images. Further, the method involves generating a flight log that includes the localized feature. Also, the method involves detecting a failure involving the navigation system, and responsively operating the camera to capture a post-failure image. The method also involves identifying one or more features in the post-failure image, and determining a location of the UAV based on a relationship between an identified feature and a localized feature.

Abstract: A method for segmenting roads in images implemented in an electronic device includes obtaining a road image; obtaining a plurality of regions of interest based on the road image; generating a splicing region based on the plurality of regions of interest; and inputting the splicing region into a pre-trained road segmentation model, and obtaining a road segmentation image and a segmentation result in the road segmentation image.

Abstract: Techniques for video manipulation detection are described to detect one or more manipulations present in digital content such as a digital video. A detection system, for instance, receives a frame of a digital video that depicts at least one entity. Coordinates of the frame that correspond to a gaze location of the entity are determined, and the detection system determines whether the coordinates correspond to a portion of an object depicted in the frame to calculate a gaze confidence score. A manipulation score is generated that indicates whether the digital video has been manipulated based on the gaze confidence score. In some examples, the manipulation score is based on at least one additional confidence score.

Abstract: An electronic device is provided.

Abstract: Embodiments include systems and methods for generating a 3D mesh on a server from one or more images and other data of a physical space that is captured by a device. The device may capture images with RGB data that includes depth information for pixels of the one or more images. The server may then create or update a 3D mesh based upon the images and depth information. The server may also use received images to provide textures to the various surfaces of the 3D mesh. The server may then cause a 3D model of the physical space to be viewed by another device, either by transmitting the 3D mesh for the other device used to construct a 3D model, or by rendering the 3D model directly on the other device. Other embodiments may be described and/or claimed.

Abstract: The present disclosure is directed to human presence detection with an infrared sensor. The human presence detection utilizes signal regularization and edge detection to minimize the effect of drift on the human presence detection. The human presence detection is accurate regardless of changes in ambient temperatures.

Abstract: Embodiments of the present disclosure relate to generating RADAR (RAdio Detection And Ranging) point clouds based on RADAR data obtained from one or more RADAR sensors disposed on one or more ego-machines. In these or other embodiments, the RADAR point clouds may be used to generate map data. Additionally or alternatively, the RADAR point clouds may be used for performing localization.

Abstract: A method for training a model for classifying videos as real or fake can include generating image tiles and audio data segments from an input video, generating a sequence of image embeddings based on the image tiles using a visual encoder and a sequence of audio embeddings based on the audio data segments using an audio encoder, transforming, using a V2A network, a first subset of the sequence of image embeddings into synthetic audio embeddings, transforming, using an A2V network, a first subset of the sequence of audio embeddings into synthetic image embeddings, updating the sequence of image embeddings by using the synthetic image embeddings, updating the sequence of audio embeddings using the synthetic audio embeddings, training the encoders and the networks using the updated sequences of image embeddings and audio embeddings, and training a classifier using the trained encoders and the trained networks.

Abstract: A method for matching a digital road map to a digital reference road map, wherein the two road maps in each case include semantic information about objects represented on the two road maps. The method includes: associating objects on the road map with objects on the reference road map on the basis of the respective semantic information such that only objects with the same semantic information are associated with one another; matching the road map to the reference road map on the basis of the association, wherein the matching comprises creating a delta pose between the road map and the reference road map. A device, a computer program, and a machine-readable storage medium are also described.

Abstract: Some aspects of the technology described herein employ a neural network with an efficient and lightweight architecture to perform spoken language recognition. Given an audio signal comprising speech, features are generated from the audio signal, for instance, by converting the audio signal to a normalized spectrogram. The features are input to the neural network, which has one or more convolutional layers and an output activation layer. Each neuron of the output activation layer corresponds to a language from a set of language and generates an activation value. Based on the activations values, an indication of zero or more languages from the set of languages is provided for the audio signal.

Abstract: Disclosed are augmented reality (AR) personalization systems to enable a user to edit and personalize presentations of real-world typography in real-time. The AR personalization system captures an image depicting a physical location via a camera coupled to a client device. For example, the client device may include a mobile device that includes a camera configured to record and display images (e.g., photos, videos) in real-time. The AR personalization system causes display of the image at the client device, and scans the image to detect occurrences of typography within the image (e.g., signs, billboards, posters, graffiti).

Abstract: An image capturing apparatus capable of performing pattern recognition with high accuracy without decreasing customer satisfaction and performing image capturing control according to a capturing scene based on a result of the pattern recognition. The image capturing apparatus includes an image sensor that includes a photoelectric conversion section including first type pixels that receive visible light from a light flux from an object and output a visible light image and second type pixels that receive invisible light from the light flux and output an invisible light image, a memory device that stores instructions, and a processor that executes the instructions to perform pattern recognition of the invisible light image, and to switch an image capturing control for capturing the visible light image from a first image capturing control to a second image capturing control when a specific pattern is detected from the invisible light image through the pattern recognition.

Abstract: Embodiments utilize pattern recognition to generate notifications in connection with analytical applications. A dashboard of the analytical dashboard is scanned to intake charts of data therefrom. Images of the charts are created, and then matched with repository patterns of a trained deep transfer model (such as a Convolutional Neural Network model). Upon matching of a pattern by the model, an alert is generated and communicated to a user to indicate a trend in the analytical data. In this manner, embodiments automatically detect data trends based upon their visual appearance when plotted in a chart, rather than through resource-intensive analysis of individual data point values. In specific embodiments, the pattern recognition may be implemented by an in-memory database engine of an in-memory database responsible for storing charts and/or chart images and/or the repository. In some embodiments, recognition of patterns in chart images may be implemented by a service.

Abstract: Embodiments described herein provide a system for three-dimensional (3D) object detection. The system includes an input interface configured to obtain 3D point data describing spatial information of a plurality of points, and a memory storing a neural network based 3D object detection model having an encoder and a decoder. The system also includes processors to perform operations including: encoding, by the encoder, a first set of coordinates into a first set of point features and a set of object features; sampling a second set of point features from the first set of point features; generating, by attention layers at the decoder, a set of attention weights by applying cross-attention over at least the set of object features and the second set of point feature, and generate, by the decoder, a predicted bounding box among the plurality of points based on at least in part on the set of attention weights.

Abstract: Aspects of the disclosure provide for mechanisms for identification of objects in images using neural networks. A method of the disclosure includes: obtaining an image, representing each element of a plurality of elements of the image via an input vector of a plurality of input vectors, each input vector having one or more parameters pertaining to visual appearance of a respective element of the image, providing the plurality of input vectors to a first subnetwork of a neural network to obtain a plurality of output vectors, wherein each of the plurality of output vectors is associated with an element of the image, identifying, based on the plurality of output vectors, a sub-plurality of elements of the image as belonging to the image of the object, and determining, based on locations of the sub-plurality of elements, a location of an image of an object within the image.

Abstract: A method for geometric calibration of a volume imaging apparatus disposes calibration phantom in a radiation path that includes a subject positioned between an x-ray source and a detector. The phantom has a number of radio-opaque markers formed of a marker material. In a repeated sequence, at each of a number of positional relationships of the x-ray source to the detector: 2D projection image data is acquired for the subject and the phantom, wherein the 2D projection image data distinguishes at least first and second x-ray energy distributions; source-to-detector geometry of the imaging apparatus is calculated, corresponding to the acquired 2D projection image data for the first and second x-ray energy distributions. The method reconstructs and displays a 3D volume image of the subject according to acquired anatomy image data from the subject and source-to-detector geometry within the 2D projection images.

Abstract: A method for converting speech in one of a plurality of input languages into text using machine transliteration and transfer learning is disclosed. The method includes a training stage. The training stage includes receiving a training set of a plurality of audio files and an input text corresponding to the audio input in any input language using the speech recognition engine; transliterating the training set to transform the input text into transliterated text that includes characters of a base language and training acoustic model with the plurality of audio files and corresponding transliterated text using transfer learning. The method further includes an inference stage. The inference stage includes performing decoding on output of the trained acoustic model to generate text includes characters of the base language at inference and transliterating the generated text to output text includes characters in input language using reverse transliteration.

Abstract: An automated registration method of 3D facial scan data and 3D volumetric medical image data using deep learning, includes extracting scan landmarks from the 3D facial scan data using a convolutional neural network, extracting volume landmarks from the 3D volumetric medical image data using the convolutional neural network and operating an initial registration of the 3D facial scan data and the 3D volumetric medical image data using the scan landmarks and the volume landmarks.

Abstract: Systems and methods for editing and generating digital ink. The present technology may provide systems and methods for training a handwriting model to generate digital ink that is stylistically and visually consistent with an original handwriting input, but which incorporates one or more changes to the text of the original handwriting input. In some examples, training may be performed using training examples that include an original handwriting sample and an original label representing the sequence of characters in the original handwriting sample. In such a case, the original handwriting sample may be processed to generate a style vector that is randomly masked, and the handwriting model may then be trained to generate a predicted handwriting sample that closely matches the original handwriting sample using the masked style vector and the original label as inputs.

Abstract: A wine label recognition method, a wine information management method and apparatus, a computer device, and a computer-readable storage medium are provided. The method includes: obtaining a wine image, and performing optical character recognition (OCR) on the wine image in a preset OCR manner, to obtain text included in the wine image (S21); performing deep learning recognition on the wine image in a preset deep learning recognition manner, to obtain an image feature included in the wine image (S22); and sifting out a target wine label matching the text and the image feature from a preset wine label database according to the text and the image feature, and using the target wine label as a wine label corresponding to the wine image (S33). Advantages of deep learning and OCR are fully utilized thereby improving accuracy and efficiency of wine label recognition and improving automation efficiency of wine information management.

Abstract: An optical sensor of the present disclosure detects light from a sheet on which is printed at least one type of single-color ink among n types of single-color inks, the optical sensor including: a light source; a light receiver including first to (n?1)th light-receiving elements; a memory configured to store a correction value based on reference data obtained by receiving lights emitted from the first to n-th single-color inks individually in the light receiver by type of single-color inks; and a controller configured to correct detection data with the correction value, the detection data obtained by receiving, in the light receiver, light emitted from a sheet irradiated with light from the light source.

Abstract: A computing device harmonizes output image properties displayed on a graphical user interface (GUI) when a user switches between two or more images of different image formats to display on the GUI. When a user interacts with a user interface to change one or more display properties of an image, a system modifies displayed image properties based on the user's modifications. If the user selects another image with different dynamic range than the first image, the system performs a backend image conversion so that the displayed properties of the second image correspond to the user-selected modifications to the displayed properties of the first image.

Abstract: Certain aspects involve a system, computer-implemented method, and computer-readable medium for identifying attributes associated with a target entity such as a person. A hierarchical characterization system receives an attribute and a request for associated identity data. The system generates an identity graph that includes attribute nodes corresponding to respective attributes and online interaction nodes corresponding to respective online interactions. The system correlates at least a subset of the online interactions and at least a subset of the attributes with a particular entity. The system generates a report indicating an identity of the entity and a behavior of the entity based on the correlated online interactions and the correlated attributes.

Abstract: Systems, methods, and other embodiments associated with computer deepfake detection are described. In one embodiment, a method includes converting audio-visual content of a person delivering a speech into a set of time series signals. Residual time series signals of residuals that indicate an extent to which the time series signals differ from machine learning estimates of authentic delivery of the speech by the person are generated. Residual values from one synchronous observation of the residual time series signals are placed into an array of residual values for a point in time. A sequential analysis of the residual values of the array is performed to detect an anomaly in the residual values for the point in time. In response to detection of the anomaly, an alert that deepfake content is detected in the audio-visual content is generated.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for adjusting a target neural network using automatically generated test cases before deployment of the target neural network in a deployment environment. One of the methods may include generating a plurality of test inputs by using a test case generation neural network; processing the plurality of test inputs using a target neural network to generate one or more test outputs for each test input; and identifying, from the one or more test outputs generated by the target neural network for each test input, failing test inputs that result in generation of test outputs by the target neural network that fail one or more criteria.

Abstract: A computer-implemented method includes identifying behavioral characteristics of a user related to operation of a computing device prior to an online account session. The method includes comparing the behavioral characteristics to a behavioral profile previously developed based on prior behavioral characteristics of the user, and determining an appropriate mark difficulty level based on a variation between the behavioral characteristics and the behavioral profile. The method includes selecting, from a plurality of prompts stored in a prompt database, a prompt that comprises instructions to draw a mark having the appropriate mark difficulty level, where other prompts of the plurality of prompts comprise instructions to draw other marks different from the mark, and providing the prompt to the user as part of a logon process for the online account session.

Abstract: This disclosure describes techniques for identifying users that are enrolled for use of a user-recognition system and updating identification data of these users over time. The user-recognition system may include a user-recognition device having projector(s) and/or camera(s) aimed substantially downwards. The device may include a projector that projects a visible light ring downwards, which the user may use to place their upward-facing palm in the appropriate location. For instance, the user may place their palm face-up with the light ring substantially centered thereon, after which a camera of the user-recognition device may generate image data of the palm, which in turn may be used to identify the user.

Abstract: A method for training an object detection model includes: firstly obtaining M sample image sets; then, obtaining an initial object detection model; and finally, training the initial object detection model by using the M sample image sets to obtain the object detection model. A sample image set includes at least one sample image and object type(s) of object(s) in each sample image. An object type corresponds to one sample image set, and the M sample image sets correspond to N object types.

Abstract: A distance measuring apparatus is provided. The apparatus comprises an image acquisition unit that acquires time-series images by an imaging unit, and a processing unit that performs image processing. The processing unit generates a plurality of enlarged images obtained by enlarging a reference image acquired before a target image among the time-series images at a plurality of types of different enlargement rates. The processing unit obtains a difference between each of the plurality of enlarged images and the target image, detects a target object that is a subject from the target image, and specifies a distance to the detected target object on the basis of the enlargement rate of the enlarged image with which the difference is minimum.

Abstract: A messaging system performs neural network hair rendering for images provided by users of the messaging system. A method of neural network hair rendering includes processing a three-dimensional (3D) model of fake hair and a first real hair image depicting a first person to generate a fake hair structure, and encoding, using a fake hair encoder neural subnetwork, the fake hair structure to generate a coded fake hair structure. The method further includes processing, using a cross-domain structure embedding neural subnetwork, the coded fake hair structure to generate a fake and real hair structure, and encoding, using an appearance encoder neural subnetwork, a second real hair image depicting a second person having a second head to generate an appearance map. The method further includes processing, using a real appearance renderer neural subnetwork, the appearance map and the fake and real hair structure to generate a synthesized real image.

Abstract: A method at a server for container location verification within a container yard, the method including requesting image data from at least one image sensor apparatus affixed to a container within the container yard; receiving the image data; and processing the image data to identify a location of a target container.

Abstract: Disclosed herein is a system and method for improving the accuracy of an object detector when trained with a dataset having a significant number of missing annotations. The method uses a novel Background Recalibration Loss (BRL) which adjusts the gradient direction according to its own activation to reduce the adverse effect of error signals by replacing the negative branch of the focal loss with a mirror of the positive branch when the activation is below a confusion threshold.

Abstract: There is provided a method of controlling a robot, which is performed by the robot that travels within a space. The robot identifies a person to be rescued, among persons within a space, by using a camera while travelling within the space, determines a distance between the robot and the person to be rescued based on depth information that is obtained by the camera and that is associated with the person to be rescued, and transmits information including the determined distance to a control server for the space.

Abstract: A method for locating an optical code, on the basis of a finder pattern of the optical code, in an image having a plurality of pixels, the method comprising the steps of determining at least one candidate position for the finder pattern on the basis of edge transitions along at least one line through the image; determining a final position of the finder pattern in a vicinity of the candidate position; and verifying the final position, i.e. verifying that the image comprises the finder pattern in the final position, using a known property of the finder pattern.

Abstract: Devices, methods, and systems for occupancy detection are described herein. One device includes instructions to receive an image of a portion of a facility captured by an imaging device, the image defined by a field of view, and set a first occupancy detection threshold for a first part of the field of view and a second occupancy detection threshold for a second part of the field of view.

Abstract: Systems and methods of authorizing access to access-controlled environments are provided. In one example, a method includes receiving, passively by a computing device, user behavior authentication information indicative of a behavior of a user of the computing device, comparing, by the computing device, the user behavior authentication information to a stored user identifier associated with the user, calculating, by the computing device, a user identity probability based on the comparison of the user behavior authentication information to the stored user identifier, receiving, by the computing device, a request from the user to execute an access-controlled function, and granting, by the computing device, the request from the user responsive to determining that the user identity probability satisfies a first identity probability threshold associated with the access-controlled function.

Abstract: A method and system for fish identification based on body surface texture features and geometric features are provided. The method employs an improved Resnet network and a deep learning Yolov8 network to extract body surface texture features and geometric features of a fish on the basis of considering influences of fish tail swing and an oxygen concentration on a fish body form based on a small sample learning framework, and then realizes identity recognition of a fish individual by coupled analysis of the body surface texture features and the geometric features. The method can realize high-accuracy fish identification with few training samples of a fish to be identified from the perspective of actual application, provides theoretical basis and technical support for accurate fish stock assessment and accurate estimation of industrially farmed fish biomass, and meets the development requirements of modern agriculture.

Abstract: A feature calculation means calculates N features for first data and N features for second data by using N feature functions for obtaining a feature for data on the basis of the data. A similarity degree calculation means calculates a similarity degree between the first data and the second data on the basis of the N features for the first data and the N features for the second data. Values of N features obtained when the same data is substituted into the N feature functions are different from each other.

Abstract: A smart table display system includes a display surface, an input/output device, and a processing circuit configured to receive financial data regarding a financial health of a user, generate a graphical user interface comprising a graphic that depicts the financial health of the user based on the financial data where the graphical user interface is for display on the display surface, receive sensor data regarding the user via the input/output device, continuously update the financial health of the user based on the sensor data, and cause the graphical user interface to update as the financial health of the user is updated.

Abstract: The invention relates to a method implemented by a computing device for processing math and text in handwriting, comprising: identifying symbols by performing handwriting recognition on a plurality of strokes; classifying, as a first classification, first symbols as either a text symbol candidate or a math symbol candidate with a confidence score reaching a first threshold; classifying, as a second classification, second symbols other than first symbols as either a text symbol candidate or a math symbol candidate with a respective confidence score by applying predefined spatial syntactic rules; updating or confirming, as a third classification, a result of the second classification by establishing semantic connections between symbols and comparing the semantic connections with the result of the second classification; and recognising each symbol as either text or math based on a result of said third classification.

Abstract: A scan driving circuit includes shift registers and clock signal lines. A shift register includes: an output circuit electrically connected to a scan input signal terminal and a pull-up node; a black frame insertion circuit electrically connected to a first clock signal terminal, a black frame insertion input signal terminal, a first voltage signal terminal, a second clock signal terminal and the pull-up node; and an output circuit electrically connected to the pull-up node, a third clock signal terminal, a shift signal terminal, a fourth clock signal terminal and a first output signal terminal. The shift registers include first shift registers and second shift registers. Third and fourth clock signal terminals of a first shift register are electrically connected to a same clock signal line. Third and fourth clock signal terminals of a second shift register are electrically connected to different clock signal lines.

Abstract: The present disclosure relates to systems and methods for processing real-time video and detecting objects in the video. In one implementation, a system is provided that includes an input port for receiving real-time video obtained from a medical image device, a first bus for transferring the received real-time video, and at least one processor configured to receive the real-time video from the first bus, perform object detection by applying a trained neural network on frames of the received real-time video, and overlay a border indicating a location of at least one detected object in the frames. The system also includes a second bus for receiving the video with the overlaid border, an output port for outputting the video with the overlaid border from the second bus to an external display, and a third bus for directly transmitting the received real-time video to the output port.

Abstract: A multi-layered moir target, useful in the calculation of misregistration between at least first, second and third layers being formed on a semiconductor device wafer, including at least one group of periodic structure stacks, each including a first stack, including a first stack first periodic structure (S1P1) having an S1P1 pitch along a first axis, a second stack, including a second stack first periodic structure (S2P1) having an S2P1 pitch along a second axis and a third stack, including a third stack first periodic structure (S3P1) having an S3P1 pitch along a third axis, the first axis being parallel to an x-axis or a y-axis, and at least one of the stacks including a second periodic structure having a second periodic structure pitch along at least one fourth axis parallel to the first axis and co-axial with one of the axes.

Abstract: The present disclosure relates to using end-to-end differentiable pipeline for optimizing parameters of a base procedural material to generate a procedural material corresponding to a target physical material. For example, the disclosed systems can receive a digital image of a target physical material. In response, the disclosed systems can retrieve a differentiable procedural material for use as a base procedural material in response. The disclosed systems can compare a digital image of the base procedural material with the digital image of the target physical material using a loss function, such as a style loss function that compares visual appearance. Based on the determined loss, the disclosed systems can modify the parameters of the base procedural material to determine procedural material parameters for the target physical material. The disclosed systems can generate a procedural material corresponding to the base procedural material using the determined procedural material parameters.

Abstract: Methods, apparatus, systems, and articles of manufacture are disclosed to decode receipts based on neural graph architecture. An apparatus includes interface circuitry to obtain an image of a document; machine readable instructions; and programmable circuitry to execute the machine readable instructions to at least generate nodes for a feature graph based on features extracted from text boxes, the nodes including polar coordinates indicative of angular positions of the text boxes; pass the feature graph through a graph neural network to generate an adjacency matrix; and identify text lines in the image of the document by clustering the nodes by line based on the adjacency matrix.

Abstract: An image processing system performs tilt correction with respect to a document image having handwritten characters and typed letters mixed with each other. The image processing system separates the document image into an image with handwritten characters determined as handwritten characters and an image without handwritten characters not determined as handwritten characters, estimates a tilt angle of the image without handwritten characters, and corrects the document image on the basis of the tilt angle.

Abstract: An apparatus comprising an interface and a processor. The interface may be configured to receive pixel data corresponding to a monitored area. The processor may be coupled to the interface and may be configured to (i) process the pixel data arranged as video frames, (ii) perform computer vision operations to detect an object in the video frames, (iii) extract characteristics about the object detected, (iv) generate a three-dimensional (3D) map of the object utilizing the pixel data corresponding to the object, (v) determine whether the object is within a three-dimensional (3D) virtual security zone based on the 3D map of the object, and (vi) generate a trigger signal in response to all or a portion of the object being within the 3D virtual security zone. The 3D virtual security zone is generally determined by a user during a setup procedure.

Abstract: The technologies described herein are generally directed to monitoring the configuration of an inventory of devices using captured images of a site. For example, a method described herein can include identifying a group of equipment installed at locations within a site. The method can further include, based on an image captured at the site by image capture equipment, detecting configuration activity at the site. Further, the method includes, based on analysis of the image, associating, by the tracking equipment, the configuration activity with equipment of the group of equipment.

Abstract: From an input image comprising a portion of text, the portion of text is extracted. The portion of text comprises text in a first natural language. The input image is classified into a candidate scene within a set of predefined scenes. The candidate scene is scored according to a quality measure. Using a predefined scene parameter of the scored candidate scene, the portion of text is translated. In the input image, the portion of text is replaced. with the translated portion of text.

Abstract: Provided herein is an integrated circuit including, in some embodiments, a special-purpose host processor, a neuromorphic co-processor, and a communications interface between the host processor and the co-processor configured to transmit information therebetween. The special-purpose host processor is operable as a stand-alone host processor. The neuromorphic co-processor includes an artificial neural network. The co-processor is configured to enhance special-purpose processing of the host processor through the artificial neural network. In such embodiments, the host processor is a keyword identifier processor configured to transmit one or more detected words to the co-processor over the communications interface. The co-processor is configured to transmit recognized words, or other sounds, to the host processor.