Abstract: A content aware and style aware neural network based data augmentation model generates augmented data sets to train neural network based handwriting recognition models to recognize individuals' handwriting. The augmented data sets may be generated so as to be artificial, and to lack personal or confidential information. The data augmentation model may generate content reference sets of individual characters generated in different fonts, and style reference sets of pluralities of characters of a particular style, for example, an individual's handwriting.

Abstract: Voltage supply system with boost converter and charge pump. A voltage supply system can include a boost converter controllable to receive an input voltage at an input node and generate an output voltage when the output voltage is greater than or equal to the input voltage. The voltage supply system can include a charge pump controllable to receive the input voltage at the input node and generate the output voltage when the output voltage is less than the input voltage. The voltage supply system can further include a controller configured to receive a control signal and control the boost converter or the charge pump to generate the output voltage at an output node based on the control signal.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for training an image embedding model. In one aspect, a method comprises: obtaining training data comprising a plurality of training examples, wherein each training example comprises: an image pair comprising a first image and a second image; and selection data indicating one or more of: (i) a co-click rate of the image pair, and (ii) a similar-image click rate of the image pair; and using the training data to train an image embedding model having a plurality of image embedding model parameters.

Abstract: A keyboard file verification method based on image processing comprises controlling a processor to perform following operations: obtaining a keyboard file; generating, according to the keyboard file, a search index and a feature image; obtaining a template image from a template database according to the search index; performing a calibration operation according to the feature image, wherein the calibration operation comprises: adjusting a resolution of the feature image according to a resolution of the template image; performing a shifting operation according to the feature image, to generate a plurality of candidate images; and comparing a key block of each of the plurality of candidate images with a key block of the template image to generate a difference map and a comparison result.

Abstract: Described herein are systems, methods, and instrumentalities associated with processing medical chest images such as chest X-ray (CXR) images. Segmentation models derived via a deep learning process are used to segment the chest images and obtain a rib segmentation result and a lung segmentation result for each image. The rib segmentation result may include a rib sequence identified in the image while the lung segmentation result may include one or more lung fields identified in the image. The quality of each chest image (e.g., whether the image reflects a breath-holding state of the patient) may then be determined based on whether a sufficient number of ribs in the rib segmentation result overlap with the lung fields in the lung segmentation result. The segmentation results may be obtained in a coarse-to-fine manner, e.g., by first determining a large rib area and then further segmenting the large rib area to identify each individual rib.

Abstract: There is provided an information processing apparatus that includes a display controlling section that causes a display device to start displaying a first virtual object while a user is moving in a real space, performs distance keeping control to substantially keep a distance between the user and the first virtual object, ends the distance keeping control in a case where a predetermined condition is satisfied, and changes the distance between the user and the first virtual object.

Abstract: A system and method for training a neural network is implemented for detecting at least one entity in a document to derive relevant inferences therefrom. The method describes obtaining at least one document. The at least one document is processed, via a detection module, to detect a widget entity. The detected widget entity is classified as active or inactive based on a detected state of the widget entity. The classified widget entity is modified into a corresponding machine-readable widget-entity based on the detected state. The at least one document is processed, via an extraction module, to detect a text entity in near vicinity of the classified widget entity. A training pair comprising the machine-readable widget entity and the corresponding text entity is generated. The neural network is trained using the generated training pair.

Abstract: The method includes receiving a plurality of digital images each depicting a tissue sample; splitting each of the received images into a plurality of tiles; automatically generating tile pairs, each tile pair having assigned a label being indicative of the degree of similarity of two tissue patterns depicted in the two tiles of the pair, wherein the degree of similarity is computed as a function of the spatial proximity of the two tiles in the pair, wherein the distance positively correlates with dissimilarity; and training a machine learning module—MLM—using the labeled tile pairs as training data to generate a trained MLM, the trained MLM being configured for performing an image analysis of digital histopathology images.

Abstract: A method and an apparatus for protecting privacy of an ophthalmic patient and a storage medium are disclosed. The method includes: acquiring an examination video of the ophthalmic patient; extracting image features of each frame of images in the examination video, performing region division on each frame of the images according to position information of organs in the image features, and obtaining a region to be precisely reconstructed and a region to be roughly reconstructed according to a region division result; performing three-dimensional (3D) reconstruction on the region to be precisely reconstructed and the region to be roughly reconstructed, to obtain 3D reconstruction data corresponding to each frame of the images; and rendering all the 3D reconstruction data into a 3D reconstruction video.

Abstract: A method is provided. The method includes generating a set of candidate training data based on a simulated environment and a first set of environmental parameters. The method also includes training a machine learning model based on the set of candidate training data. The method further includes obtaining a set of segmentations based on the machine learning model and a set of test data. The method further includes determining whether a mean intersection-over-union (MIOU) of the set of segmentations has increased by more than a threshold change. The method further includes generating a next set of candidate training data based on the simulated environment and a second set of environmental parameters for the simulated environment, in response to determining that the MIOU has increased by more than the threshold MIOU amount.

Abstract: A generative adversarial network-based classification system and method that can generate missing data as missing data imputation values similar to real data using a generative adversarial network (GAN) and allowing training with labeled data sets with labels, as well as and irregular data sets such as non-labeled data sets without labels.

Abstract: Disclosed is method for dynamic cropping of full body pose (FBP) image, method comprising: receiving in computing device FBP image from image resource, wherein FBP image comprises face; detecting in received FBP image area of face by bounding box; detecting coordinates of corner points of bounding box; performing distance approximation of detected area of face in received FBP image; cropping distance approximated area of face of FBP image; performing coordinates mapping transformation of cropped FBP image by mapping coordinates of cropped FBP image to received FBP image; segmenting region of hair of cropped FBP image; masking region of hair of cropped FBP image for obtaining hair polymask; and applying masked FBP image to received FBP image according to performed coordinates mapping for obtaining the dynamically cropped FBP image.

Abstract: Multi-planar detection and imaging systems and methods are provided. The system has a radio device, at least one antenna, and a processing system in communication with the radio device. The radio device includes at least one receiver and a first transmitter transmitting radio waves in multiple planes. The antenna transmits the radio waves in multiple frequencies. A second transmitter transmits a carrier radio wave guiding the radio waves such that the radio waves identify points of interest in an object or living entity and reflect back to the radio device. The receiver receives the reflected radio waves, and the processing system analyzes the points of interest and constructs an image of the object or living entity. An AI engine including a recurrent neural network may be provided, and the recurrent neural network controls transmission of the radio waves.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for generating descriptions of input images. One of the methods includes obtaining an input image; processing the input image using a first neural network to generate an alternative representation for the input image; and processing the alternative representation for the input image using a second neural network to generate a sequence of a plurality of words in a target natural language that describes the input image.

Abstract: An image recognition method includes acquiring a target image, the target image being an image of a certificate to be recognized; text area recognition is performed on the target image to obtain a text area image of a target text corresponding to the certificate to be recognized; a text direction of the target text is determined according to the text area image; direction adjustment is performed on the text area image according to the text direction to obtain an adjusted text area image; and text recognition is performed on the adjusted text area image to obtain a text content of the target text.

Abstract: A prediction system harvests geo-tagged ground-level images through one or more algorithms. The system receives point of interest data representing structures or events and tags the geo-tagged ground-level images with a probability describing a classification. The system tags point of interest data with a hierarchical genre classification and encodes the tagged geo-tagged ground-level images as vectors to form nodes and edges in a proximity graph. The system encodes tagged points of interest data as similarity vectors to render more nodes and more edges on the proximity graph associated with the tagged geo-tagged ground-level images nodes by calculated semantic distances. The system splits the proximity graph into a training subgraph and a testing subgraph and trains a neural network by aggregating and sampling information from neighboring nodes within the training subgraph graph and validates through the testing subgraph. Training ends when a loss measurement is below a threshold.

Abstract: An apparatus for optimizing workflows of one or more microscopes and/or microscope systems includes one or more processors and one or more computer-readable storage media. The one or more computer-readable storage media have stored therein computer-executable instructions, which, when executed by the one or more processors cause execution of the following steps: implementing, by one or more components of the one or more microscopes and/or microscope systems, a workflow comprising a capture of first data; applying one or more trained models to the captured first data; and making at least one decision in relation to the workflow based on the application of the one or more trained models to the captured first data.

Abstract: A radio frequency domain calibration system and method for radio frequency circuit units is provided. The system regulates and calibrates radio frequency integrated circuit devices, circuit modules and transceiver circuit systems including circuit components in the radio frequency domain through calibration units. The calibration unit includes a calibration control unit and a calibration regulation unit, where the calibration control unit generates a calibration control signal, and drives the calibration regulation unit to perform programmable, adaptive, and real-time regulation on the circuit structure, geometric scale, and working parameters of the radio frequency circuit unit, so that the radio frequency circuit unit works in a desired condition with optimized performances. Compared with existing digital domain calibration systems, the radio frequency domain calibration system has smaller chip area, lower cost, and higher regulation flexibility and module configuration flexibility.

Abstract: A tethered drone system has a drone that carries telecommunication equipment, such as one or more cellular transceivers and an antenna, that enables the system to provide access to a telecommunication network for a plurality of remote communication devices in a vicinity of the drone. The drone may be disassembled to facilitate transport to a location at which additional cellular service is desired, such as a remote location or an area impacted by a catastrophic weather or geological event. At or near such location, the drone may be quickly assembled and launched. While airborne, the drone may hover for an extended period of time, such as several, hours, days, weeks, or even longer, while the telecommunication equipment carried by the drone provides access to the telecommunication network.

Abstract: A pixel-level object detection system that detects an object at a pixel level, includes: an imaging unit that acquires an inference image that is an image that captures a detection target; an area detection unit that detects an area including a detection target from the inference image; a detail detection unit that detects the detection target using only local information from the inference image; and a result integration processing unit that integrates an output of the area detection unit with an output of the detail detection unit to output a segmentation map indicating, as a probability map, which pixel in the image corresponds to the detection object.