Abstract: Examples of a computer-implemented regularization technique are described and tailor-made for Vision Transformers. Examples can combine two or more input images and swap patches between them to produce a PatchSwap image. The PatchSwap image is then used to train a Vision Transformer to predict the mixing ratio as well as the categories of the original images.

Abstract: Examples of computer-implemented training techniques are described and tailor-made for Vision Transformers (ViTs). Example techniques include training a model (network) for the rotation of images and image patches and training of the model to predict the rotation angles. The model learns to extract both global and local features from an image.

Abstract: Exemplary embodiments of wearable stretch sensors and applications of using the same are disclosed. In embodiments, the sensors and the applications disclosed herein can be used to capture spinal motion and posture information.

Abstract: Described herein are means for integrating a defense mechanism into deep-learning-based systems to defend against adversarial attacks. For instance, an exemplary system is specially configured for adding a convolutinal defense layer to a neural network containing orthogonal kernels. Such a system generates the convolutional defense layer based on generating a set of learned kernals to increase diversity of network architecture, in which generating the set of learned kernals includes feeding an output of the convolutional defense layer into the neural network, further in which generating the convolutional defense layer includes selecting one or more orthogonal kernals, duplicating as needed and arranging them in a particular order.

Abstract: Systems and methods for a content-adaptive photo-enhancement recommender that produces multiple enhanced images based on content and color layout parameters derived from related high-quality images are disclosed. In some embodiments, a user-uploaded image is enhanced based on parameters learned from high-quality images similar in content to the user-uploaded image and obtained using content-based image retrieval techniques.

Abstract: Systems and methods for a content-adaptive photo-enhancement recommender that produces multiple enhanced images based on content and color layout parameters derived from related high-quality images are disclosed. In some embodiments, a user-uploaded image is enhanced based on parameters learned from high-quality images similar in content to the user-uploaded image and obtained using content-based image retrieval techniques.

Abstract: A method, apparatus and computer program product are provided for identifying an unknown subject using face recognition. In particular, upon receiving a plurality of images depicting a subject, the method may include deriving and storing a common component image and a gross innovation component image associated with the subject, wherein the subject can later be identified in a new image using these two stored images. The common component image may capture features that are common to all of the received images depicting the subject, whereas the gross innovation component image may capture a combination of the features that are unique to each of the received images. The method may further include deriving and storing a low-rank data matrix associated with the received images, wherein the low-rank data matrix may capture any illumination differences and/or occlusions associated with the received images.

Abstract: An intelligent system that supports real-time and offline feedback based on automated analysis of a trainee's performance using data streams captured in the training process is disclosed.

Abstract: A system including a video stream and a data stream. The system synchronizes the data stream to different portions of the video stream.

Abstract: The selection of favorites for a video presentation system.

Abstract: A method of selecting first and second basis views from a group of at least three provided views where the first and second basis views are used in synthesizing a third virtual view. The group of provided views are spatially offset from one another in at least one dimension.

Abstract: Summarization of video content, such as a video that includes sumo, by a method that provides a description of the video, identifies a plurality of segments of the video based upon the provided description, and generates another video based upon the identified segments, having less frames than the original video.

Abstract: Summarization of video content including football.

Abstract: A system for presenting a summarization of audio and/or visual content having a plurality of segments to a user together with a graphical user interface that preferably indicates to the viewer the relative temporal position of video segments viewed in the summary within the content from which the summary was derived.

Abstract: A system including a video stream and a data stream. The system synchronizes the data stream to different portions of the video stream.

Abstract: A system including a video stream and a data stream. The system synchronizes the data stream to different portions of the video stream.

Abstract: Summarization of video content including soccer.

Abstract: Identification of segments based upon scoreboards and logos.

Abstract: Summarization of video content including football.

Abstract: A system for modifying images to be shown on displays that have display characteristics dependant on the angle at which a displayed image is viewed. An image may be modified by detecting the position of a viewer relative to a display and based on the detected position, correcting the image.