Abstract: In various examples, three-dimensional (3D) object or feature detection and localization for autonomous and semi-autonomous systems and applications is described herein. Systems and methods are disclosed that use different types of sensors, such as an image sensor and a LIDAR sensor, to determine information associated with objects, such as traffic objects (e.g., traffic signs, traffic signals, traffic markings, etc.). To determine the information for an object, image data is processed to determine a bounding shape associated with the object. The bounding shape is then used to determine a 3D shape, such as a frustum, corresponding to the object. Additionally, points data generated using the LIDAR sensor, such as an occupancy map and/or a point cloud, is processed to identify a portion of the points associated with (e.g., located within) the 3D shape. This portion of the points may then be used to determine the information associated with the object.

Abstract: A method performed by a processor of a computing system is described herein, where the method includes obtaining an image that includes an object having a shape, where a boundary of the shape of the object in the digital image is labeled in the digital image. The method also includes computing an encoding for the shape, where computing the encoding for the shape includes partitioning the shape into multiple partitions. Computing the encoding for the shape further includes, for the multiple partitions, computing angle-based contour descriptors that represent boundaries of the partitions, where the encoding for the shape of the object is based upon the angle-based contour descriptors.

Abstract: The embodiments of the present disclosure provide a CCD camera calibration system, method, device, computing device, and storage medium. The system includes: a transport mechanism; a calibration component disposed on the transport mechanism and provided with multiple calibration pattern units arranged in a length direction of the calibration component, the calibration pattern unit including a first calibration block, which includes a positioning identifier and multiple first calibration patterns uniformly arranged; a CCD camera for photographing the calibration component to obtain a first image while the calibration component is travelling; a processing module, which is in communication connection with the CCD camera, and is adapted to calculate a transformation matrix of internal and external parameters of the CCD camera based on the first image.

Abstract: A computer-implemented method for obtaining a combined image from two images, the computer-implemented method comprising: obtaining a first image and a second image of a biometric feature of a user, the biometric feature comprising a plurality of characteristics; identifying, in the first image and in the second image, the characteristics; storing the first and second image and the identified characteristics; for at least some of the identified characteristics in the first image, identifying corresponding identified characteristics in the second image; deriving a transformation for transforming the at least some of the identified characteristics in the second image into the corresponding identified characteristics in the first image; and modifying at least a portion of the second image using the transformation and obtaining a combined image comprising at least one portion of the first image and the at least one modified portion of the second image.

Abstract: This disclosure provide an apparatus which comprises an acquisition unit that acquires an image, a divider that divides the image into a plurality of partial images; a converter that converts the partial images into tokens having fixed dimensional vectors; an adder that adds a class-token having a vector of a dimension same as the two or more tokens to the tokens obtained by the converter; an encoder that updates the tokens obtained by the adder based on a relevance between the tokens to obtain final encoded representations; an acquiring unit that acquires an encoded representation corresponding to the class-token in the encoded representations obtained by the encoder as class-token encoded representations; and a combining unit that combines the class-token encoded representations to obtain a feature vector of the input image.

Abstract: Systems, apparatus, and methods for augmenting dense depth maps using sparse data. Various embodiments combine single-image depth estimation (SIDE) techniques with structure-from-motion techniques for improved depth accuracy. In some examples, a machine learning (ML) model is used to generate a dense depth map based on one or more frames/images of a video. Structure-from-motion (SfM) analysis is performed on the video to determine depth information from camera movement in the video. The structure-from-motion techniques may generate more accurate data than the ML model, however, the data from the ML model may be denser compared with the SFM depth data. The dense ML model depth map may be augmented by the SFM depth data. Augmentation may include fitting the relative depths determined by the ML model depth map to the absolute depth in the SFM data resulting in more accurate dense depth information.

Abstract: The disclosure provides example visual localization methods and devices. One method includes that a terminal device obtains an image of a building. The terminal device generates a descriptor based on the image. The descriptor includes information about a horizontal viewing angle between a first vertical feature line and a second vertical feature line in the image. The first vertical feature line indicates a first facade intersection line of the building, and the second vertical feature line indicates a second facade intersection line of the building. The terminal device performs matching in a preset descriptor database based on the descriptor, to obtain localization information of a photographing place of the image.

Abstract: A medical image processing device includes a display device; and an electronic processor programmed to perform a medical image processing method including receiving a three dimensional (3D) medical image; computing least one 3D derivative image of the 3D medical image; identifying a set of fibers in the 3D medical image by tracing fibrous image features in the at least one 3D derivative image starting from respective seed locations in the least one 3D derivative image; and controlling the display device to display an anatomical representation comprising or derived from the set of fibers.

Abstract: A transformer-based machine-learned model may predict object behavior by using cross-attention between dynamically-sized patches of a top-down representation of the environment and an object location and/or previous behavior. A patch's size may be based at least in part on map data and/or sensor data and a portion of the top-down representation associated with an area of the environment that is outside a roadway may be excluded from patch generation.

Abstract: This disclosure provides systems, methods, and devices for vehicle driving assistance systems that support image processing. In a first aspect, a method of image processing includes receiving image data from an image sensor; receiving ranging data from a ranging sensor; embedding first spatial features of the image data with first temporal information associated with the image data; embedding second spatial features of the ranging data with second temporal information associated with the ranging data; determining first bird's-eye-view (BEV) features based on the first spatial features embedded with first temporal information; determining second BEV features based on the second spatial features embedded with second temporal information; and determining, based on the first and second BEV features, a feature set for processing by a transformer network. The feature set includes at least a portion of both the first and second BEV features. Other aspects and features are also claimed and described.

Abstract: A layout of a robotic operational environment is optimized to at least reduce or minimize occlusion of one or more sensors, to improve computational efficiency of motion planning, and/or improve an efficiency and/or robustness of generated motions plans. A respective sensor pose (i.e., positioned and oriented) for one or more sensors may be determined while taking into account the various robot poses of the robot(s) and transitions between robot poses. For instance, cost values that represent occlusion may be determined, and candidate sensor poses evaluated based on cost values or aggregated cost. The total number of sensors to be employed may not be known before execution, and in fact may be determined only after several iterations, upon reaching a termination condition (e.g., task completion time). Sensors make take any variety of forms, e.g., digital camera, radar, LIDAR, sonar, passive infrared (PIR) motion detector.

Abstract: An electronic device and a control method of an electronic device are provided. The method acquiring a plurality of images through at least one camera, inputting red green blue (RGB) data for each of the plurality of images into a first neural network model to obtain two-dimensional pose information on an object included in the plurality of images, inputting RGB data for at least one image of the plurality of images into a second neural network model to identify whether the object is transparent, if the object is a transparent object, performing stereo matching based on the two-dimensional pose information on each of the plurality of images to obtain three-dimensional pose information on the object, and if the object is an opaque object, acquiring three-dimensional pose information on the object based on one image of the plurality of images and depth information corresponding to the one image.

Abstract: An information processing method is provided. The method includes obtaining a target video. A first target image feature corresponding to a face image of each frame is obtained. A target identity coefficient and a target texture coefficient corresponding to the target image feature in the face image of a different frame in the target video are obtained. A first target identity feature is obtained according to the target identity coefficient, and a first target texture feature is obtained according to the target texture coefficient. Once a first target feature is obtained by splicing the target image feature, the first target identity feature and the first target texture feature, a first target expression coefficient is obtained based on the first target feature.

Abstract: Embodiments of the disclosure discloses method and apparatus for facial beauty prediction, device and storage medium. The method includes the steps of classifying a training set into a plurality of first images with noise labels and a plurality of second images with non-noise labels; performing a re-weighting processing on the plurality of the second images; training a first model by using a target training set to obtain a second model; labeling image data by using the second model to obtain first data with labels and second data without labels; generating third data with pseudo labels through a classifier by using the second data; training the classifier according to the first data, the second data and the third data; processing an image to be predicted through a model with a target classifier to obtain a facial beauty prediction result.

Abstract: Systems and methods for image segmentation are provided. The systems may obtain a target image and a template image relating to the target image. The template image may correspond to an initial mask reflecting initial segmentations of the template image. The systems may determine a first transformation and an intermediate template image by preliminarily registering the template image to the target image and generate an intermediate mask based on the initial mask and the first transformation. The systems may determine, based on the intermediate mask, one or more first regions from the target image and one or more second regions from the intermediate template image. The systems may determine a second transformation by registering each of the one or more second regions to a corresponding first region. The systems may determine a target mask according to which the target image can be segmented based on one or more second transformations.

Abstract: The technology of this application relates to a pose determining method. The method includes obtaining a first image, when first pose information determined based on the first image meets a pose anomaly condition, displaying prompt information for indicating to photograph a target object, and obtaining, by using the target object in the target image captured by a user based on the prompt, second pose information that does not meet the pose anomaly condition. When high-precision pose information cannot be determined, pose positioning is implemented by using the target object in a scenario. In addition, in a process of determining pose information of a terminal device, prompt information for indicating the user to photograph the target object is displayed, and the user is indicated to photograph the target object, thereby avoiding that the user does not know how to operate or scan an invalid target object.

Abstract: Disclosed herein are an immersive image encoding/decoding method and apparatus, and a method for transmitting a bitstream generated by the immersive image encoding method. An immersive image encoding method according to the present disclosure may include: generating geometry atlases for an immersive image; aligning the geometry atlases to make an alignment height of the geometry atlases correspond to a height of a texture atlas; and generating a geometry bitstream by encoding the aligned geometry atlases.

Abstract: The present invention is directed to a monitoring apparatus and method for preventing pressure injuries. More specifically, an apparatus and method that is portable, multi-functional, and easy to use, capable of passive monitoring, AI triage monitoring, and EMR integration that can detect deep tissue injuries, falls, and/or bed wetness, using an equitable approach to detection.

Abstract: A detection system includes an acquisition unit, a map generation unit, a determination unit, and a detection unit. The acquisition unit acquires a first image and a second image. The second image is an image acquired by capturing an image of a narrower area than in the first image. The map generation unit generates, by using the first image, map information. The determination unit determines a target area in the map information. The detection unit analyzes the second image acquired by capturing an image of the target area, and thereby detects a target object.

Abstract: Conditional procedural model generation techniques are described that enable generation of procedural models that are usable to recreate a visual appearance of an input image. A content processing system, for instance, receives a training dataset that includes a plurality of training pairs. The content processing system trains a machine learning model to generate procedural models based on input images. The content processing system then receives an input image that has a particular visual appearance. The content processing system leverages the trained machine learning model to generate a procedural model that is usable to recreate the particular visual appearance of the input digital image.