Abstract: A system and method of calibrating a broadcast video feed are disclosed herein. A computing system retrieves a plurality of broadcast video feeds that include a plurality of video frames. The computing system generates a trained neural network, by generating a plurality of training data sets based on the broadcast video feed and learning, by the neural network, to generate a homography matrix for each frame of the plurality of frames. The computing system receives a target broadcast video feed for a target sporting event. The computing system partitions the target broadcast video feed into a plurality of target frames. The computing system generates for each target frame in the plurality of target frames, via the neural network, a target homography matrix. The computing system calibrates the target broadcast video feed by warping each target frame by a respective target homography matrix.

Abstract: The disclosure provides a system that may receive an identification of a first patient; may receive a first template that includes first multiple locations associated with a face of the first patient and associated with the identification of the first patient; may determine second multiple locations associated with a face of a current patient; may determine a second template of the face of the current patient based at least on the second multiple locations associated with the face of the current patient; may determine if the first template matches the second template; if the first template matches the second template, may provide an indication that the current patient has been correctly identified as the first patient; and if the first template does not match the second template, may provide an indication that the current patient has not been identified.

Abstract: A learning model learned to provide high image quality of a first image is generated. A first image and a second image are received from the same target, high image quality of the first image is provided by using the learned model, and a first high image quality image is obtained. By using the first high image quality image and the second image as inputs, a high image quality image of the second image having the image quality of the first high image quality image is generated while maintaining contrast of the second image.

Abstract: A computer-implemented method of handling an emergency incident can include receiving information on an emergency incident that includes at least one image of the emergency incident, applying a Convolutional Neural Network (CNN) object recognition and classification process for identifying and marking objects on the at least one image that are related to the emergency incident and that may cause at least one secondary hazardous situation, processing the data relating to the identified and marked objects by applying a deep learning algorithm to the data in a Relational Network (RN) architecture, wherein the image on the basis of the identified and marked objects is correlated to a set of recognized objects in a database for classifying the emergency. A communication network, communication apparatus, and an emergency processing unit are also provided. Embodiments of such machines and systems can be configured to implement embodiments of the method.

Abstract: An image identification system includes: an image dividing unit which divides training images of a first training data set and assigns a label assigned to a training image from which dividing occurs to the divided training images as tentative labels; a texture index computing unit which computes texture indexes for each of the divided training images; a tentative label prediction model building unit which builds a tentative label prediction model to predict tentative labels assigned to the divided training images based on the texture indexes; and a label comparison unit which compares first tentative labels assigned to the divided training images with second tentative labels predicted with respect to the divided training images by the tentative label prediction model and extracts divided training images for which there is discrepancy between the first and second tentative labels as those images for which it is necessary to modify tentative labels.

Abstract: Systems and methods are described for performing a computer vision task on a sequence of frames. A first frame and a second frame are obtained, corresponding to a first timestep and a second timestep, respectively, in a sequence of frames. A differential image is computed between the first frame and the second frame. A predicted output is generated by forward propagating the differential image through a neural network that is trained to perform a computer vision task.

Abstract: This application discloses an artificial intelligence-based object detection method and apparatus. The method includes inputting a target image comprising an object to an object detection model; obtaining feature images of different scales from the target image using the object detection model; determining image location information of the object and a first confidence level that the object belongs to each category; acquiring a target region in which the object is located; inputting the target region to an object retrieval model, comparing the target region with sample images of a plurality of categories to obtain a second confidence level that the object belongs to each category; and determining a target category of the object based on the first confidence level and the second confidence level, a sum of the first confidence level and the second confidence level of the target category being the largest of the plurality of categories.

Abstract: A processor acquires a first radiographic image and a second radiographic image obtained by irradiating a subject, into which a tracheal tube has been inserted, with radiation in different directions, derives a three-dimensional position of a distal end of the tracheal tube in the subject on the basis of a position of the distal end of the tracheal tube in each of the first radiographic image and the second radiographic image, derives a three-dimensional position of a bronchial bifurcation in the subject on the basis of a bronchial bifurcation position in each of the first radiographic image and the second radiographic image, and determines insertion of the tracheal tube into an esophagus on the basis of the three-dimensional position of the distal end of the tracheal tube and the three-dimensional position of the bronchial bifurcation.

Abstract: An electronic device and method are disclosed. The electronic device includes a memory and processor, which implements the method, including: obtaining image data for an image, set in the image a region of interest and a background region, using depth information, both including one or more sub-regions, determining respective bokeh characteristics for the one or more first sub-regions and the one or more second sub-regions based on a first characteristic criterion and a second characteristic criterion, respectively, and processing the image to apply a bokeh effect to the plurality of pixels based on the respective bokeh characteristics for the one or more first sub-regions and the one or more second sub-regions.

Abstract: An intelligent multi-visual camera system is disclosed that includes a multiple visual sensor array and a control system. The multiple visual sensor array includes multiple visual cameras spaced apart from each other. The control system also receives input from the multiple visual cameras, stores instructions that cause the processor to: initiate a registration system that projects images from the multiple visual cameras into a single three-dimensional volume and coordinate overlapping pixels of adjacent images amongst each other; detect one or more features of interest in the images from the multiple visual cameras; track one or more features of interest in one frame in one image from the multiple visual cameras into a subsequent frame in a subsequent image from the multiple visual cameras; deduplicate the projected images from the multiple visual cameras onto the ground plane; and communicate information regarding the features of interest that have been detected.

Abstract: A footstep planning method includes: obtaining a number of depth images of an environment in a walking direction of a legged robot; creating a three-dimensional model of the environment based on the depth images; determining at least one even region from the three-dimensional model of the environment; and selecting one or more of the at least one even region as one or more candidate footstep locations for the legged robot to step on.

Abstract: A testing system for an image processing algorithm including a control unit, an image processing device, an image processing hardware, and a testing device is disclosed. The control unit provides an original image and parameters. The image processing device obtains the original image and the parameters, and drives the image processing hardware to perform a first image processing procedure to the original image based on the parameters to generate a hardware-processed image. The testing device obtains the original image, the parameters, and the hardware-processed image, and performs, through a simulation software, a second image processing procedure to the original image based on the parameters to generate a software-processed image. The testing device compares the hardware-processed image with the software-processed image through a testing software to generate a comparing result that shows a pixel difference of the hardware-processed image and the software-processed image.

Abstract: Aspects of the present invention relate to a control system (2) for monitoring a movement of a trailer (3) connected to a vehicle (1). The control system comprises one or more controllers (5) and is configured to: receive image data from an image sensor (10) disposed on the vehicle; store image data corresponding to a central reference image, a current reference image and a current frame and associate an offset angle with the current reference image, and associate a correction angle with current reference image. The control system (2) is configured to compare the current reference image to the current frame to determine a change angle; determine a hitch angle in dependence on the offset angle, the change angle and the correction angle; and output the hitch angle.

Abstract: A method of encoding point cloud data comprises storing, in a buffer, a maximum coordinate of a pair of coordinates of an applicable node, wherein the applicable node is a most-recently encoded node with a same position as a current node along an applicable axis and the pair of coordinates are for axes different from the applicable axis; determining a context for a planar mode plane position of the current node, wherein determining the context for the planar mode plane position comprises determining, based on the maximum coordinate of the pair of coordinates of the applicable node, a distance value representing a distance between the current node and the applicable node; determining an increment value that indicates whether the distance value is greater than a threshold; and determining the context index based on the increment value; and entropy encoding the planar mode plane position using the determined context.

Abstract: Techniques related to multi-level optical flow estimation are discussed. Such techniques include partitioning each pair of input images into one or more partitions, separately performing optical flow estimation on the partitions, and merging the separately generated optical flow results into a final optical flow map for the pair of input images.

Abstract: A method according to one embodiment of the present disclosure comprises receiving a first image of a patient's anatomy, the first image generated at a first time and depicting a plurality of rigid elements; receiving a second image of the patient's anatomy, the second image generated at a second time after the first time and depicting the plurality of rigid elements; determining a transformation from the first image to the second image for each one of the plurality of rigid elements to yield a set of transformations; calculating a homography for each transformation in the set of transformations to yield a set of homographies; and identifying, using the set of homographies, a common portion of each transformation attributable to a change in camera pose, and an individual portion of each transformation attributable to a change in rigid element pose.

Abstract: Techniques for identity preserved controllable facial image manipulation are described that support generation of a manipulated digital image based on a facial image and a render image. For instance, a facial image depicting a facial representation of an individual is received as input. A feature space including an identity parameter and at least one other visual parameter is extracted from the facial image. An editing module edits one or more of the visual parameters and preserves the identity parameter. A renderer generates a render image depicting a morphable model reconstruction of the facial image based on the edit. The render image and facial image are encoded, and a generator of a neural network is implemented to generate a manipulated digital image based on the encoded facial image and the encoded render image.

Abstract: There is provided a method and apparatus of training a model, a device, and a medium, which relate to artificial intelligence, and in particular to a deep learning and image processing technology. The method may include: determining a plurality of augmented sample sets associated with a plurality of original samples; determining a first constraint according to a first model based on the plurality of augmented sample sets; determining a second constraint according to the first model and a second model based on the plurality of augmented sample sets, wherein the second constraint is associated with a difference between outputs of the first model and the second model for one augmented sample, and the first model has a complexity lower than that of the second model; training the first model based on at least the first constraint and the second constraint, so as to obtain a trained first model.

Abstract: An image processing device includes an encoder configured to receive a blurry image and generate a global feature map of the image, a merging unit configured to merge the global feature map and blur information, a decoder configured to generate a feature tensor and weight tensors, a recurrent refinement module configured to perform recurrent feature filtering, and an image reconstruction module configured to reconstruct a deblurred image where the image processing device is configured to estimate an image global shift and to activate or deactivate the recurrent refinement module based on the estimation.

Abstract: A digital video camera architecture for updating an object identification and tracking model deployed with the camera is disclosed. The invention comprises optics, a processor, a memory, and an artificial intelligence logic which may further comprise artificial neural networks. The architecture identifies objects according to a first confidence threshold of the model and identifies candidate objects according to the first confidence threshold and a second confidence threshold. The model may track the motion of the candidate objects within a visual field, separate the candidate objects into false positive candidate objects and false negative candidate objects according to their tracked motions, and present at least a portion of the false positive candidate objects and false negative candidate objects for further annotation.