Abstract: Systems and methods are provided for multi-modal test-time adaptation. The method includes inputting a digital image into a pre-trained Camera Intra-modal Pseudo-label Generator, and inputting a point cloud set into a pre-trained Lidar Intra-modal Pseudo-label Generator. The method further includes applying a fast 2-dimension (2D) model, and a slow 2D model, to the inputted digital image to apply pseudo-labels, and applying a fast 3-dimension (3D) model, and a slow 3D model, to the inputted point cloud set to apply pseudo-labels. The method further includes fusing pseudo-label predictions from the fast models and the slow models through an Inter-modal Pseudo-label Refinement module to obtain robust pseudo labels, and measuring a prediction consistency for the pseudo-labels.

Abstract: An information processing apparatus (20) according to the present disclosure includes a control unit (230). When a display image is displayed on a transmissive display in which real space is visually recognizable, a control unit (230) detects, from the display image, a transparent area through which the real space is seen. The control unit (230) corrects pixel values of at least a part of an area in the transparent area of the display image.

Abstract: A method and an apparatus for decoding an image are disclosed. A region of the image is selected and the decoding is selected region and associated metadata is performed. Pixels for a generated for a decoded image based on the decoded selected region and metadata.

Abstract: A computing system can perform computer-vision-based operations to create a vision-based user environment. The system can process an image stream and identify a computer vision character of a vision alphabet in the image stream. The vision alphabet includes a collection of computer vision characters, with each computer vision character having a distinctive characteristic identifiable by computer vision. The system can process an image of the video stream to provide to a user device a function associated with the identified computer vision character, the processing including to make a bounded space within the video stream corresponding to the identified computer vision character interactive on the user device. The system can provide the function in response to interaction with the bounded space at the user device. The system can provide a processed image to the user device, or simply provide functionality for a received stream.

Abstract: Systems and methods of the present disclosure can facilitate determining a three-dimensional surface representation of an object. In some embodiments, the system includes a computer, a calibration module, which is configured to determine a camera geometry of a set of cameras, and an imaging module, which is configured to capture spatial images using the cameras. The computer is configured to determine epipolar lines in the spatial images, transform the spatial images with a collineation transformation, determine second derivative spatial images with a second derivative filter, construct epipolar plane edge images based on zero crossings of second derivative epipolar planes image based on the epipolar lines, select edges and compute depth estimates, sequence the edges based on contours in a spatial edge image, filter the depth estimates, and create a three-dimensional surface representation based on the filtered depth estimates and the original spatial images.

Abstract: The present disclosure provides methods and systems for vehicle verification. The method may include receiving vehicle verification information related to a vehicle to be verified from a client, wherein the vehicle verification information includes a plurality of images acquired by the client via an imaging device, and the vehicle verification information responds to at least one vehicle verification instruction. The method may further include determining a verification result of the vehicle based on the vehicle verification information.

Abstract: A vision transformer network having extremely low latency and usable on mobile devices, such as smart eyewear devices and other augmented reality (AR) and virtual reality (VR) devices. The transformer network processes an input image, and the network includes a convolution stem configured to patch embed the image. A first stack of stages including at least two stages of 4-Dimension (4D) metablocks (MBs) (MB4D) follow the convolution stem. A second stack of stages including at least two stages of 3-Dimension MBs (MB3D) follow the MB4D stages. Each of the MB4D stages and each of the MB3D stages include different layer configurations, and each of the MB4D stages and each of the MB3D stages include a token mixer. The MB3D stages each additionally include a multi-head self attention (MHSA) processing block.

Abstract: A system and method for processing candidate edges are presented. The method may be performed when spatial structure information and 2D image information are stored. The method may include identifying, based on the spatial structure information, a 3D location that represents a corner of an object structure; identifying, based on the spatial structure information, a 3D vector that extends from the 3D location and is parallel with an edge; determining a 2D location within the 2D image information that corresponds to the 3D location; determining a 2D vector within the 2D image information that corresponds to the 3D vector; determining an edge detection region; identifying a cluster of candidate edges within the edge detection region that does not represent any object edge; identifying, as a 2D object edge, a candidate edge that is not part of the identified cluster; and performing segmentation based on the 2D object edge.

Abstract: Provided is a method for acquiring images. The method includes: acquiring a first image embodying a display panel; determining repair indication information of a plurality of defects in the display panel in the first image, wherein the repair indication information indicates a degree of a repair demand for the defects; selecting a target defect from the plurality of defects based on the repair indication information of the plurality of defects; and acquiring a repair reference image embodying the target defect by photographing the target defect in the display panel.

Abstract: Methods for creating personalized items using images associated with a subject are provided. The method includes receiving and processing at least one image of a subject; translating the processed at least one image into a personalized illustration having features that substantially resemble features of the subject, wherein translating the processed at least one image comprises extracting a plurality of core shapes from a face of the subject in the at least one image and assembling the extracted plurality of core shapes into a face of the personalized illustration; and manipulating the personalized illustration of the subject for placement in a personalized item such that the manipulated personalized illustration appears as an integral part of an existing template of the personalized item.

Abstract: A device for locating a coupling means arranged on a trailer vehicle before the coupling operation is described, having a coupling on the towing vehicle side that is designed with an insertion opening which is open in the rearward direction (R) and is intended for the entry of the coupling means, a light transmitter fitted to the coupling on the towing vehicle side, and a light receiver which captures light from the light transmitter that is reflected by the coupling means. The problem of providing a device which can be used to make it possible for the coupling on the towing vehicle side to more precisely approach the coupling means of the trailer vehicle is solved by virtue of the fact that the light transmitter is a stroboscopic light.

Abstract: In order to enhance enhanced X-ray image inhalation quality monitoring, a metric is proposed hat reproducibly provides an index of ribs visible to be used in the assessment of the inhalation state. In an example, a detected diaphragm in a chest X-ray image may be projected into an atlas that contains labels for all intercostal spaces, namely spaces between rib centerlines. A spatial representation of both the clavicle and the ribs is provided in the atlas, a cumulative histogram is built for all points, i.e. pixels, of the diaphragm, for every point a rib label counter of the rib in the rib label map at that point is incremented as well as all ribs above it, the rib label counter is normalized by a division by the number of points, the median (or a different quantile) may be taken of this distribution serving as an inhalation index. An objective metric of inhalation state is thus achieved.

Abstract: A sensing data processing device is provided. The sensing data processing device may track keypoints within a full observable range of the sensing data processing device based on sensing data collected from a previous frame and a sensing data collected from a current frame through a reference sensor, the full observable range determined based on a combined field of view (FOV) of sensors, manage a new keypoint together with the keypoints in response to the new keypoint being detected, and perform localization of the device based on the managed keypoints.

Abstract: An image processing apparatus performs first noise reduction processing on a plurality of radiation images corresponding to mutually-different radiation energies, generates a decomposition image by energy subtraction processing using the plurality of radiation images obtained by performing the first noise reduction processing, and performs second noise reduction processing on the decomposition image, wherein the second noise reduction processing uses a filter that differs from a filter used in the first noise reduction processing in at least one of size and type.

Abstract: The present disclosure describes techniques for annotating a video. The techniques comprises receiving an operation of creating a bullet screen of annotating a target object in a first frame; determining a display position of the bullet screen in the first frame; displaying the bullet screen at the display position in the first frame; determining an offset of the display position of the bullet screen relative to a position of the target object in the first frame; and displaying the bullet screen at a target display position in a second frame of the video, wherein the target display position in the second frame is determined based at least in part on the offset of the display position of the bullet screen relative to the position of the target object in the first frame, the second frame is played subsequent to the first frame, and the second frame comprises the target object.

Abstract: A method for detecting from images correct placement and function, or incorrect placement and function, of a clip of a transportation box of wafers in sterile or similar conditions obtains an image template comprising features of clip and obtains a first detection image of a working clip. An object region focusing on the imaged clip in the first detection image is determined according to the image template. Part of the working image is selected as a first preset location. The part of the image is masked to obtain a second detection image, the masking obscures the background region of the part of the image but not the clip-object region, and displays the unobscured clip-object region. The second detection image is input into a trained neural network model to determine whether the clip is in sealed or unsealed state. An electronic device and a non-transitory storage medium are also disclosed.

Abstract: In some embodiments, a computer-implemented method includes obtaining a without-eyeglasses face scan of a subject, the without-eyeglasses face scan being a three-dimensional (3D) model of a face of the subject without eyeglasses; obtaining a with-eyeglasses face scan of the subject, the with-eyeglasses face scan being a 3D model of the subject with eyeglasses; and using the without-eyeglasses face scan and the with-eyeglasses face scan to place a 3D eyeglasses model on a face model of the subject. In some embodiments of the computer-implemented method, the 3D eyeglasses model is placed on the face model of the subject using frame placement information generated using the without-eyeglasses face scan and the with-eyeglasses face scan.

Abstract: A processor-implemented method with object pose estimation includes: determining an image feature corresponding to a point cloud of an input image; determining semantic segmentation information, instance mask information, and keypoint information of an object, based on the image feature; and estimating a pose of the object based on the semantic segmentation information, the instance mask information, and the keypoint information.

Abstract: Systems and method for an object from a plurality of objects are disclosed. An image of a scene containing the plurality of objects is obtained, and a segmentation map is generated for the objects in the scene. The shapes of the objects are determined based on the segmentation map. An end effector is adjusted in response to determining the shapes of the objects. The adjusting the end effector includes shaping the end effector according to at least one of the shapes of the objects. The plurality of objects is approached in response to the shaping of the end effector, and one of the plurality of objects is picked with the end effector.

Abstract: Example aspects include techniques for implementing a high-frequency attention network for single image super-resolution. These techniques may include extracting a plurality of features from an original image input into a CNN to generate a feature map, and restoring one or more high-frequency details of the original image via an efficient residual block (ERB) and a high-frequency attention block (HFAB) configured to assign a scaling factor to one or more high-frequency areas. In addition, the techniques may include generating reconstruction input information by performing an element-wise operation on the one or more high-frequency details and cross-connection information from the feature map and performing, by the CNN, an enhancement operation on the reconstruction input information to generate an enhanced image.