Abstract: A cascaded machine learning inference system and method is disclosed. The cascaded system and method may be designed to be employed in resource restricted environments. The cascaded system and method may be applicable for applications that operate with limited power (e.g., a wearable smart watch). The cascaded system and method may employ two or more subsystems that are operable to classify an input signal provided by any number or types of sensors suitable for a given application. For instance, the sensors used may include gyroscopes, accelerometers, magnetometers, or barometric altimeters. The system and method may also be further split functionality across additional or new subsystems. By splitting operations and functionality across additional subsystems, the overall power consumption may further be reduced.

Abstract: A prism of an approximately quadrangle-frustum shape is arranged so that a bottom side, out of two parallel surfaces of the prism, is a placing surface side for a finger. A first imaging unit arranged below a top surface parallel to the bottom surface images an image of the finger transmitted through the top surface. A light source radiates light to at least one side surface of a first set of side surfaces, out of two sets of side surfaces of the approximately quadrangle-frustum shape that face each other. A second imaging unit images the image of the finger transmitted through a second set of side surfaces, out of the two sets of side surfaces. An infrared ray light source radiates infrared ray light into the finger so that the infrared ray light is scattered inside the finger and is received by the imaging unit.

Abstract: Tracking units for facial features with advanced training for natural rendering of human faces in real-time are provided. An example device receives a video stream, and upon detecting a visual face, selects a 3D model from a comprehensive set of head orientation classes. The device determines modifications to the selected 3D model to describe the face, then projects a 2D model of tracking points of facial features based on the 3D model, and controls, actuates, or animates hardware based on the facial features tracking points. The device can switch among an example comprehensive set of 35 different head orientation classes for each video frame, based on suggestions computed from a previous video frame or from yaw and pitch angles of the visual head orientation. Each class of the comprehensive set is trained separately based on a respective collection of automatically marked images for that head orientation class.

Abstract: A method, a computer program product, and a computer system identifies a last person of a queue. The method includes receiving an image of the queue where the image includes a plurality of individuals comprising the queue. The method includes determining positions and facing directions of the individuals comprising the queue. The method includes identifying the last person of the queue based on a vector field analysis according to the positions and the facing directions of the individuals comprising the queue. The method includes generating instructions to join the queue based on the identified last person of the queue. The instructions include a modified image indicating the last person of the queue in the image.

Abstract: A method includes receiving image data representing an environment of a robotic device from a camera on the robotic device. The method further includes applying a trained dense network to the image data to generate a set of feature values, where the trained dense network has been trained to accomplish a first robot vision task. The method additionally includes applying a trained task-specific head to the set of feature values to generate a task-specific output to accomplish a second robot vision task, where the trained task-specific head has been trained to accomplish the second robot vision task based on previously generated feature values from the trained dense network, where the second robot vision task is different from the first robot vision task. The method also includes controlling the robotic device to operate in the environment based on the task-specific output generated to accomplish the second robot vision task.

Abstract: A device, method and system for estimating elevation in images from camera devices is provided. The device detects humans at respective positions in images from a camera device, the camera device having a fixed orientation and fixed focal length. The device estimates, for the humans in the images, respective elevations of the humans, relative to the camera device, at the respective positions based at least on camera device parameters defining the fixed orientation and the fixed focal length. The device associates the respective elevations with the respective positions in the images. The device determines, using the respective elevations associated with the respective positions, a function that estimates elevation in an image from the camera device, using a respective image position coordinate as an input. The device provides the function to a video analytics engine to determine relative real-world positions in subsequent images from the camera device.

Abstract: An information processing apparatus includes a frame image acquisition section adapted to acquire a plurality of consecutive frame images included in a moving image displayed on a screen, and a matching process section adapted to perform, for each of the plurality of acquired frame images, a matching process of detecting an area that matches a template image representing appearance of a display element to be detected. An area in which the display element is being displayed on the screen is identified on a basis of a result of performing the matching process on the plurality of frame images.

Abstract: Method and system for wild animal deterrence, the method includes obtaining video data by a monitor camera of a home wildlife deterrence system; classifying, based on the obtained video data, an object in the video data as a particular type of a wild animal; selecting an action to perform based on the particular type of the wild animal that the object is classified as; and triggering the action to be performed. The method also includes determining that the particular type of the wild animal matches a label of a candidate action in a set of candidate actions, wherein each candidate action in the set of candidate actions indicates at least one type of wild animal; and in response to determining that the particular type of the wild animal matches a label of the candidate action in the set of candidate actions, selecting the candidate action as the action to perform.

Abstract: Image modification using detected symmetry is described. In example implementations, an image modification module detects multiple local symmetries in an original image by discovering repeated correspondences that are each related by a transformation. The transformation can include a translation, a rotation, a reflection, a scaling, or a combination thereof. Each repeated correspondence includes three patches that are similar to one another and are respectively defined by three pixels of the original image. The image modification module generates a global symmetry of the original image by analyzing an applicability to the multiple local symmetries of multiple candidate homographies contributed by the multiple local symmetries. The image modification module associates individual pixels of the original image with a global symmetry indicator to produce a global symmetry association map.

Abstract: A method and an apparatus for recognizing an object are disclosed. The apparatus may extract a plurality of features from an input image using a single recognition model and recognize an object in the input image based on the extracted features. The single recognition model may include at least one compression layer configured to compress input information and at least one decompression layer configured to decompress the compressed information to determine the features.

Abstract: In certain embodiments, the systems, apparatus, and methods disclosed herein relate to robotic surgical systems with built-in navigation capability for patient position tracking and surgical instrument guidance during a surgical procedure, without the need for a separate navigation system. Robotic based navigation of surgical instruments during surgical procedures allows for easy registration and operative volume identification and tracking. The systems, apparatus, and methods herein allow re-registration, model updates, and operative volumes to be performed intra-operatively with minimal disruption to the surgical workflow. In certain embodiments, navigational assistance can be provided to a surgeon by displaying a surgical instrument's position relative to a patient's anatomy.

Abstract: Determining that a motor vehicle driver is using a mobile device while driving a motor vehicle. Multiple images of a driver of a motor vehicle are captured through a side window of the motor vehicle. Positive images show a driver using a mobile device while driving a motor vehicle. Negative images show a driver not using a mobile device while driving a motor vehicle. Multiple training images are selected from both the positive images and the negative images. The selected training images and respective labels, indicating that the selected training images are positive images or negative images, are input to a machine (e.g. Convolutional Neural Network, (CNN)). The CNN is trained to classify that a test image, captured through a side window of a motor vehicle, shows a driver using a mobile device while driving the motor vehicle.

Abstract: Implementations include receiving a first intermediate document representation of a first document, the first intermediate document representation including OCR data of document text and positions of document text in the first document, receiving a set of intermediate template representations for a set of templates, the set of intermediate template representations including, for each template in the set of templates, OCR data representative of template text and positions of template text in a respective template, determining a first sub-set of templates using text-based detection based on the first intermediate document representation and the set of intermediate template representations, determining a second sub-set of templates using position-based detection based on the first intermediate document representations and intermediate template representations associated with one or more templates in the first sub-set of templates, and transmitting instructions to execute template-based data extraction for the fi

Abstract: Examples of the present disclosure provide a face matching method and a face matching apparatus, and a storage medium. The face matching method includes: obtaining a first attribute of first face information which is to be matched; determining one or more preferential matching ranges based on the first attribute; and comparing the first face information with second face information in the one or more preferential matching ranges.

Abstract: A device for and computer implemented method of image content recognition and of training a neural network for image content recognition. The method comprising collecting a first set of digital images from a database, the first set of digital images is sampled from digital images assigned to a many shot class; creating a first training set comprising the collected first set of digital images; training a first artificial neural network comprising a first feature extractor and a first classifier for classifying digital images using the first training set; collecting first parameters of the trained first feature extractor, collecting second parameters of the trained classifier, determining third parameters of a second feature extractor of a second artificial neural network depending on the first parameters, determining fourth parameters of a second classifier for classifying digital images of the second artificial neural network.

Abstract: Methods, systems, and programming for user identification are presented. In one example, a system for acquiring biometric information is disclosed. The system comprises a housing including a surface for a person to place a finger thereon. The system also comprises a sensor, a first image acquisition portion, and a second image acquisition portion. The sensor is configured for sensing presence of the finger when the person places the finger on the surface. The first image acquisition portion is configured for acquiring a fingerprint image of the finger placed on the surface. The second image acquisition portion is configured for acquiring a finger vein image of the finger placed on the surface. The first and second image acquisition portions acquire their respective images at different times.

Abstract: A three-dimensional position estimation device includes: a feature point extracting unit for detecting an area corresponding to the face of an occupant in an image captured by a camera for imaging a vehicle interior and extracting a plurality of feature points in the detected area; an inter-feature-point distance calculating unit for calculating a first inter-feature-point distance that is a distance between distance-calculating feature points among the plurality of feature points; a face direction detecting unit for detecting the face direction of the occupant; a head position angle calculating unit for calculating a head position angle indicating the position of the head of the occupant with respect to an imaging axis of the camera; an inter-feature-point distance correcting unit for correcting the first inter-feature-point distance to a second inter-feature-point distance that is a distance between distance-calculating feature points in a state where portions of the head corresponding to the distance-calcu

Abstract: Methods, systems, and programming for identifying a person are presented. In one example, a first set of candidates is determined based on a fingerprint image acquired from a finger of the person. A second set of candidates is determined based on a finger vein image acquired from the finger. The person is recognized based on the first and/or second set of candidates.

Abstract: A system and method for classifying images of an evidence, the method including: generating an unlabeled template based on a first received image, wherein the unlabeled template includes a first set of regions of interest (ROIs); selecting a labeled template that includes a second set of ROIs from a plurality of labeled templates determined to be a best match for the unlabeled template; checking at least an ROI of the first set of ROIs for content that corresponds to a label of an equivalent ROI of the second set of ROIs; and, classifying the first image upon determination that the content checked corresponds to the label according to the labeled template.

Abstract: Methods and systems for developing a sizing system through categorization and selection of prototypes, which can be regarded as the most appropriate fit model, is described. Once categorized and prototypes are selected, recommendations for the sizing of a target body part may be issued.