The present invention provides a method for quantitatively identifying the defects of large-size composite material based on infrared image sequence, firstly obtaining the overlap area of an infrared splicing image, and dividing the infrared splicing image into three parts according to overlap area: overlap area, reference image area and registration image area, then extracting the defect areas from the infrared splicing image to obtain P defect areas, then obtaining the conversion coordinates of pixels of defect areas according to the three parts of the infrared splicing image, and further obtaining the transient thermal response curves of centroid coordinate and edge point coordinates, finding out the thermal diffusion points from the edge points of defect areas according to a created weight sequence and dynamic distance threshold ε.sub.ttr×d.sub.p_max, finally, based on the thermal diffusion points, the accurate identification of quantitative size of defects are completed.

An electronic apparatus performs a method of constructing a facial position map from a two-dimensional (2D) facial image of a real-life person that includes: generating a coarse facial position map from the 2D facial image; predicting a first set of keypoints in the 2D facial image based on the coarse facial position map; identifying a second set of keypoints in the 2D facial image based on the user-provided keypoint annotations; and updating the coarse facial position map to get a final facial position map so as to reduce the differences between the first set of keypoints and the second set of key points in the 2D facial image. In some embodiments, a final set of keypoints and/or a three-dimensional (3D) facial model of the real-life person is formed from the final facial position map.

Method and apparatus for coding system using affine motion model are disclosed. According to one method, a neighbouring block set of the current block comprising multiple spatial neighbouring blocks and one or more collocated blocks is determined for the current block. One or more constructed affine MVP candidates are derived for an affine MVP candidate list based on CP (control-point) MVs (motion vectors) at multiple spatial neighbouring blocks and said one or more collocated blocks. One constructed affine MVP candidate without one temporal MV is checked and inserted into the affine MVP candidate list before any constructed affine MV with one temporal MV. The current block or motion information of the current block is then encoded or decoded based on the affine MVP candidate list.

Methods and apparatus are disclosed for implementing machine learning data augmentation within the die of a non-volatile memory (NVM) apparatus using on-chip circuit components formed on or within the die. Some particular aspects relate to configuring under-the-array or next-to-the-array components of the die to generate augmented versions of images for use in training a Deep Learning Accelerator of an image recognition system by rotating, translating, skewing, cropping, etc., a set of initial training images obtained from a host device. Other aspects relate to configuring under-the-array or next-to-the-array components of the die to generate noise-augmented images by, for example, storing and then reading training images from worn regions of a NAND array to inject noise into the images.

A method for automatically classifying emission tomographic images includes receiving original images and a plurality of class labels designating each original image as belonging to one of a plurality of possible classifications and utilizing a data generator to create generated images based on the original images. The data generator shuffles the original images. The number of generated images is greater than the number of original images. One or more geometric transformations are performed on the generated images. A binomial sub-sampling operation is applied to the transformed images to yield a plurality of sub-sampled images for each original image. A multi-layer convolutional neural network (CNN) is trained using the sub-sampled images and the class labels to classify input images as corresponding to one of the possible classifications. A plurality of weights corresponding to the trained CNN are identified and those weights are used to create a deployable version of the CNN.

A dual-screen user device and methods for revealing a combination of selected desktops and applications on single and dual screens are disclosed. Desktops and applications can be shifted between screens by user gestures, and/or moved off of the screens and therefore hidden. Hidden desktops and screens can be re-displayed by other gestures. The desktops and applications are arranged in a window stack that represents a logical order of the desktops and applications providing a user with an intuitive ability to manage multiple applications/desktops running simultaneously. One embodiment provides an annunciator window extending across both screens in a dual screen configuration. The annunciator window provides alerts, notifications, and statuses of the device in an increased area thereby enhancing viewability of the information in the window. The annunciator window can be expanded over a selected screen to view full contents of the window without having to minimize or close running applications.

A method may be provided to operate a medical system. First data may be provided for a first 3-dimensional (3D) image scan of an anatomical volume, with the first data identifying a blood vessel node in a first coordinate system for the first 3D image scan. Second data may be provided for a second 3D image scan of the anatomical volume, with the second data identifying the blood vessel node in a second coordinate system for the second 3D image scan. The first and second coordinate systems for the first and second 3D image scans of the anatomical volume may be co-registered using the blood vessel node identified in the first data and in the second data as a fiducial.

A method for generating an object detection dataset and a computer implemented object detection system are disclosed. The method comprises:

receiving a training image dataset comprising a plurality of images that include objects of interest, each image comprising pixel values corresponding to an imaged material generated by a penetrating imager;

generating a thresholded image for each of the plurality of images;

segmenting each thresholded image into images corresponding to objects;

creating a greyscale image per object from the segmented images corresponding to that object by, for each object, calculating an average pixel value for each pixel of the object from corresponding pixels of the object in the segmented images;

forming a greyscale image for the object from the averaged pixels;

storing the greyscale images in a data repository as an object detection dataset.

The present invention provides systems and methods for tracking items (e.g., commodities, goods, containers, boxes, packages, etc.) through transportations to multiple locations to allow the position(s) and movement(s) of such items to be accurately tracked and documented, and to allow such items to be quickly identified and located based on tracking records kept within the tracking system. The system may utilize image sensors, image recognition and processes software, position translation software, and a virtual model of the pre-defined space in order to track objects within the defined space and maintain a record of the movement(s) and position(s) of the object within the pre-defined space.

Techniques for creating machine learning models for direct homography regression for image rectification are described. In certain embodiments, a training service trains an algorithm on a source view of a training image and a homography matrix of the training image into a machine learning model that generates a normalized homography matrix for an input of the source view. The normalized homography matrix may then be utilized to generate a target view of an image input into the machine learning model. The target view of the image may be used in a document processing pipeline for document images captured using cameras.