Abstract: An image recognition method and a terminal, where the method includes obtaining, by the terminal, an image file including a target object recognizing, by the terminal, the target object based on an image recognition model in the terminal using a neural network computation apparatus in the terminal to obtain object category information of the target object, and storing, by the terminal, the object category information in the image file as first label information of the target object. Hence, image recognition efficiency of the terminal can be improved, and privacy of a terminal user can be effectively protected.

Abstract: A system for recording, storing and processing diagnostic information, including: a computer implementing a computer-readable media including digital data and ground truth; a registry constructed and arranged to store and associate transactions or accesses on the data; and a machine learning system that considers each learning step modification a microtransaction for the data used in that step and which is recorded in the transaction registry. Other embodiments of this aspect include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the methods.

Abstract: Disclosed is a method including receiving, at a server and from a browser configured on a device, a request to access the server, the server being associated with a brick and mortar store, transmitting an interactive interface to the browser on the device, and receiving, from the browser on the device and via use of the interactive interface, data associated with a product in the brick and mortar store. The server transmits via a browser application programming interface that defines a protocol for communicating data between the server and the browser a request associated with a purchase of the product and receives according to the API authorized payment data. The user can scan products codes in the store using the interactive interface. The process eliminates the need for point of sale equipment and eliminates checkout lines.

Abstract: Digital image layout training is described using wireframe rendering within a generative adversarial network (GAN) system. A GAN system is employed to train the generator module to refine digital image layouts. To do so, a wireframe rendering discriminator module rasterizes a refined digital training digital image layout received from a generator module into a wireframe digital image layout. The wireframe digital image layout is then compared with at least one ground truth digital image layout using a loss function as part of machine learning by the wireframe discriminator module. The generator module is then trained by backpropagating a result of the comparison.

Abstract: A method of rendering a stereoscopic 3D image from a single image, including receiving a collection of pairs of 3D images including an input image and an output image that is a 3D pair of the input image, training a neural network composed of convolutional layers without any fully connected layers, with pairs of 3D images from the collection, to receive an input image and generate an output image that is a 3D pair of the input image, wherein the neural network is provided as an application on a computing device, receiving an input image and generating an output image that is a 3D pair of the input image by the neural network.

Abstract: Provided is a virtual reality (VR) device including a receiver configured to receive, from a three-dimensional (3D) camera being a polyhedron, images captured by cameras arranged at each vertex of the polyhedron; a memory storing the images; a processor configured to generate a complex view by synthesizing the images; and a display configured to display the complex view.

Abstract: A method including receiving a source file containing a plurality of documents which, to a computer, initially are indistinguishable from each other. A first classification stage is applied to the source file using a convolutional neural network image classification to identify source documents in the multitude of documents and to produce a partially parsed file having a multitude of identified source documents. The partially parsed file includes sub-images corresponding to the plurality of identified source documents. A second classification stage, including a natural language processing artificial intelligence, is applied to sets of text in bounding boxes of the sub-images, to classify each of the multitude of identified source documents as a corresponding sub-type of document. Each of the sets of text corresponding to one of the sub-images. A parsed file having a multitude of identified sub-types of documents is produced. The parsed file is further computer processed.

Abstract: Methods, apparatuses and electronic devices for identifying an object category include: determining M key point neighborhood regions from corresponding object candidate boxes according to position information of M key points in a plurality of object candidate boxes of an image to be detected, where M is less than or equal to the total number of key points of N preset object categories, and M and N are positive integers; and determining category information of at least one object in the image to be detected using a convolutional neural network model used for identifying an object category in the image according to the M key point neighborhood regions.

Abstract: Disclosed herein are systems, methods, and computer-readable storage devices for a new browser including multiple application programming interfaces. A method includes receiving, from a site, at a browser and via a first application programming interface that defines a first protocol for communicating data between the browser and the site, a first payment request associated with a potential purchase by a user, in response to the first payment request and based on an identification of a payment service, communicating, from the browser and via a second application programming interface that defines a second protocol for communicating data between the browser and the payment service, a second payment request to the payment service, receiving, at the browser, from the payment service, via the second application programming interface, authorized payment information and communicating, from the browser, to the site and via the first application programming interface, the authorized payment information.

Abstract: A method for training an object detection network by using attention maps is provided. The method includes steps of: (a) an on-device learning device inputting the training images into a feature extraction network, inputting outputs of the feature extraction network into a attention network and a concatenation layer, and inputting outputs of the attention network into the concatenation layer; (b) the on-device learning device inputting outputs of the concatenation layer into an RPN and an ROI pooling layer, inputting outputs of the RPN into a binary convertor and the ROI pooling layer, and inputting outputs of the ROI pooling layer into a detection network and thus to output object detection data; and (c) the on-device learning device train at least one of the feature extraction network, the detection network, the RPN and the attention network through backpropagations using an object detection losses, an RPN losses, and a cross-entropy losses.

Abstract: Techniques are disclosed for predicting the probability of successfully building the software application whether an integration build between source components of a software application will be successful. An integration service executing on a server computer determines whether to test an integration build of source components of a software application. The integration service obtains metrics related to the developer of each component as well as any previous integration builds of the components. Based on the metrics, the integration service predicts a probability of a successful integration build of the source components. Based on the probability of a successful integration build, the integration service may notify a developer to perform an action to increase the likelihood of the integration build being successful.

Abstract: A computer that identifies a fake image is described. During operation, the computer receives an image. Then, the computer performs analysis on the image to determine a signature that includes multiple features. Based at least in part in the determined signature, the computer classifies the image as having a first signature associated with the fake image or as having a second signature associated with a real image, where the first signature corresponds to a finite resolution of a neural network that generated the fake image, a finite number of parameters in the neural network that generated the fake image, or both. For example, the finite resolution may correspond to floating point operations in the neural network. Moreover, in response to the classification, the computer may perform a remedial action, such as providing a warning or a recommendation, or performing filtering.

Abstract: An offset component of multiplication by a transistor is to be reduced. An imaging device includes a pixel region, a first circuit, a second circuit, a third circuit, and a fourth circuit. The pixel region includes a plurality of pixels, and a pixel includes a first transistor. An offset potential and a weight potential are supplied to the pixel selected by the first circuit and the second circuit. The pixel obtains a first signal by photoelectric conversion with use of light. The first transistor multiplies the first signal by the weight potential. The first transistor generates a first offset term and a second offset term with use of a multiplication term of the first signal by the weight potential and the offset potential. The third circuit subtracts the first offset term, and the fourth circuit subtracts the second offset term. The fourth circuit determines the multiplication term, and the fourth circuit outputs a determination result through the neural network interface.

Abstract: Methods and systems are provided for determining high-dynamic range lighting parameters for input low-dynamic range images. A neural network system can be trained to estimate lighting parameters for input images where the input images are synthetic and real low-dynamic range images. Such a neural network system can be trained using differences between a simple scene rendered using the estimated lighting parameters and the same simple scene rendered using known ground-truth lighting parameters. Such a neural network system can also be trained such that the synthetic and real low-dynamic range images are mapped in roughly the same distribution. Such a trained neural network system can be used to input a low-dynamic range image determine high-dynamic range lighting parameters.

Abstract: The present disclosure discloses a method and apparatus for detecting a human face. A specific embodiment of the method comprises: acquiring a to-be-detected image; inputting the to-be-detected image into a pre-trained first convolutional neural network to obtain facial feature information, the first convolutional neural network being used to extract a facial feature; inputting the to-be-detected image into a pre-trained second convolutional neural network to obtain semantic feature information, the second convolutional neural network being used to extract a semantic features of the image; and analyzing the facial feature information and the semantic feature information to generate a face detection result. This embodiment improves accuracy of a detection result of a blurred image.

Abstract: Systems and methods for domain adaptation are provided. The system aligns image level features between a source domain and a target domain based on an adversarial learning process while training a domain discriminator. The system selects, using the domain discriminator, unlabeled samples from the target domain that are far away from existing annotated samples from the target domain. The system selects, based on a prediction score of each of the unlabeled samples, samples with lower prediction scores. The system annotates the samples with the lower prediction scores.

Abstract: In one embodiment, example systems and methods relate to a manner of generating training data for a classifier or a regression function using labeled synthetic images and a mapping that accounts for the differences between synthetic images and real images. The mapping may be a neural network that was trained using image pairs that each include an image of an object and a synthetic image that is generated from the image of the object by overlaying a rendering of the object into the image. The mapping may recognize the differences between features of the object in the real image and features of the rendering of the object in the synthetic image such as color, contrast, sensor noise, etc. Later, a set of labeled synthetic images is received, and the mapping is used to generate training data from the labeled synthetic images.

Abstract: A training method of a neural network, and a recognition method and apparatus using the neural network are disclosed. The recognition method using the neural network includes obtaining a feature vector generated from a hidden layer of the neural network, in response to data being entered to an input layer of the neural network, and determining a reliability of a recognition result for the data using the feature vector and clusters.

Abstract: An apparatus to facilitate compute optimization is disclosed. The apparatus includes a plurality of processing units each comprising a plurality of execution units (EUs), wherein the plurality of EUs comprise a first EU type and a second EU type.

Abstract: A method for dividing, by a training system, a computational training work load of one or more neural network layers; pre-training the one or more neural network layers with a first class of image data sensitive to an original known dataset; generating a first weight file from the first layer of the neural network based on the first class of image data sensitive to the original known dataset; loading the one or more pre-trained neural network layers and the generated first weight file into at least one Internet of Things (IoT) device; stacking the one or more pre-trained neural network layers with the first layer of the neural network to form a new training system for an uploaded new dataset; adjusting the generated first weight file based on an input of one or more new classes of image data comprised in the uploaded new dataset to generate a new second weight file; inferencing an object class of new image data comprised on the uploaded new dataset using the generated new second weight file; and outputting th

Abstract: A shift register includes a first input sub-circuit, a pull-up control sub-circuit, and a pull-down control sub-circuit. The first input sub-circuit is configured to transmit a voltage from the first signal terminal to the first node under control of the first voltage terminal. The pull-up control sub-circuit is configured to be in a turn-on or turn-off state under control of the first node. The pull-down control sub-circuit is configured to transmit a voltage from the third voltage terminal to the pull-down node under control of the first node, transmit the voltage from the third voltage terminal to the pull-down node under control of the signal output terminal, and transmit a voltage from the first clock signal terminal to the pull-down node under control of the first clock signal terminal.

Abstract: Triage of training data for acceleration of large-scale machine learning is provided. In various embodiments, training input from a set of training data is provided to an artificial neural network. The artificial neural network comprises a plurality of output neurons. Each output neuron corresponds to a class. From the artificial neural network, output values are determined at each of the plurality of output neurons. From the output values, a classification of the training input by the artificial neural network is determined. A confidence value of the classification is determined. Based on the confidence value, a probability of inclusion of the training input in subsequent training is determined. A subset of the set of training data is determined based on the probability. The artificial neural network is trained based on the subset.

Abstract: A three-dimensional model generating device includes: a converted image generating unit that, for each of input images included in one or more items of video data and having mutually different viewpoints, generates a converted image from the input image that includes fewer pixels than the input image; a camera parameter estimating unit that detects features in the converted images and estimates, for each of the input images, a camera parameter at a capture time of the input image, based on a pair of similar features between two of the converted images; and a three-dimensional model generating unit that generates a three-dimensional model using the input images and the camera parameters.

Abstract: A system of devices receives and stores documents based on confidential information redacted from the documents. An electronic document is analyzed to identify character blocks having confidential information. The confidential information can be in different formats within the document. Redaction rules are applied to the character blocks to identify confidential categories for the confidential information within the blocks. The confidential information is redacted based on the rules such that the confidential information is removed from the document. A new electronic document is generated with the information redacted such that it is not viewable or printable. The two documents with different levels of confidential information is then stored on separate devices within the system.

Abstract: A table layout determination system implemented on a computing device obtains an image of a table having multiple cells. The table layout determination system includes a row prediction machine learning system that generates, for each of multiple rows of pixels in the image of the table, a probability of the row being a row separator, and a column prediction machine learning system generates, for each of multiple columns of pixels in the image of the table, a probability of the column being a column separator. An inference system uses these probabilities of the rows being row separators and the columns being column separators to identify the row separators and column separators for the table. These row separators and column separators are the layout of the table.

Abstract: Methods and systems for selecting a selected-sub-set of features from a plurality of features for training a machine learning module, the training of the machine learning module to enable classification of an electronic document to a target label, the plurality of features associated with the electronic document. In one embodiment, the method comprises analyzing a given training document to extract the plurality of features, and for a given not-yet-selected feature of the plurality of features: generating a set of relevance parameters iteratively, generating a set of redundancy parameters iteratively and determining a feature significance score based on the set of relevance parameters and the set of redundancy parameters. The method further comprises selecting a feature associated with a highest value of the feature significance score and adding the selected feature to the selected-sub-set of features.

Abstract: In one example, a system for a sort function race can include a processor, and a memory resource storing instructions executable by the processor to generate a plurality of variant sort functions that utilize a variant of an input from a parent sort function, that perform the plurality of variant sort functions as a plurality of separate processes, identify a variant sort function from the plurality of variant sort functions that completes a function in a shortest period of time, and receive results from the identified variant sort function.

Abstract: Techniques for neural network training are provided. One computer-implemented method comprises: obtaining, by an electronic device operatively coupled to a processing unit, based on a similarity between two images, a first image and a second image. The computer-implemented method also comprises training, by the electronic device, a neural network based on the first image and the second image such that a distance between a first vector and a second vector generated respectively from the first image and the second image in the trained neural network is associated with the similarity.

Abstract: A method and an image processing entity for applying a convolutional neural network to an image are disclosed. The image processing entity processes the image while using the convolutional kernel to render a feature map, whereby a second feature map size of the feature map is greater than a first feature map size of the feature maps with which the feature kernel was trained. Furthermore, the image processing entity repeatedly applies the feature kernel to the feature map in a stepwise manner, wherein the feature kernel was trained to identify the feature based on the feature maps of the first feature maps, wherein the feature kernel has the first feature map size.

Abstract: Disclosed is a method including presenting an input field on a user interface of a generalized search entity, wherein the generalized search entity processes data using a generalized search engine that indexes and searches both merchant sites and non-merchant sites, receiving user input in the input field and determining whether the user input corresponds to a product in a product database to yield a determination. When the determination indicates that the user input does correspond to the product in the product database, the method includes presenting a purchase-related search result, wherein the purchase-related search result is configured such that when a user interacts with the purchase-related search result and confirms a purchase associated with the purchase-related search result, the generalized search entity initiates a purchasing process for the product.

Abstract: Embodiments of the invention are directed to systems, methods, and computer program products for approaching less spatial resolution documents with parameterized and evolution linear units and momentum-driven SGD to accelerate the training phase of the system further by transformation of electronic documents by low-resolution intelligent up-sampling. An up-sampling layer is integrated with dots-per-inch (DPI) to validate whether the desirable output is obtained.

Abstract: An apparatus in one embodiment comprises at least one processing device comprising a processor coupled to a memory. The processing device is configured to determine a plurality of block offsets indicating respective positions of data within a plurality of uncompressed data blocks respectively encoded to have variable sizes, to generate a block index file comprising the plurality of block offsets, a plurality of block positions and a plurality of block sizes respectively corresponding to the plurality of uncompressed data blocks, to compress data from at least one uncompressed data block of the plurality of uncompressed data blocks to at least one compressed data block of a plurality of compressed data blocks, and to utilize the block index file in connection with decompressing the compressed data.

Abstract: Disclosed herein are system, method, and computer program product embodiments for robust key value extraction. In an embodiment, one or more hierarchical concepts units (HCUs) may be configured to extract key value and hierarchical information from text inputs. The HCUs may use a convolutional neural network, a recurrent neural network, and feature selectors to analyze the text inputs using machine learning techniques to extract the key value and hierarchical information. Multiple HCUs may be used together and configured to identify different categories of hierarchical information. While multiple HCUs may be used, each may use a skip connection to transmit extracted information to a feature concatenation layer. This allows an HCU to directly send a concept that has been identified as important to the feature concatenation layer and bypass other HCUs.

Abstract: A mobile device can establish a communication with a separate device via a single function action such as bringing the devices near to each other. A method can include establishing a communication between a mobile device and a separate device a via a wireless link, presenting an instruction associated with the potential purchase and receiving, after the instruction is displayed and interpreted by the mobile device, a single-function gesture which can be a security measure to prevent unauthorized purchase. The method includes retrieving the payment data from a memory of the mobile device and transmitting the payment data via the wireless link to the separate device to make a purchase.

Abstract: System and methods for machine learning are described. A first input value is obtained. A second input value is also obtained. A decision to use for generating a cycle output is selected based on a randomness factor. The decision is at least one of a random decision or a best decision from a previous cycle. A cycle output for the first and second inputs is generated using the selected decision. The selected decision and the resulting cycle output are stored.

Abstract: Digital image defect identification and correction techniques are described. In one example, a digital medium environment is configured to identify and correct a digital image defect through identification of a defect in a digital image using machine learning. The identification includes generating a plurality of defect type scores using a plurality of defect type identification models, as part of machine learning, for a plurality of different defect types and determining the digital image includes the defect based on the generated plurality of defect type scores. A correction is generated for the identified defect and the digital image is output as included the generated correction.

Abstract: Supervised machine learning using convolutional neural network (CNN) is applied to denoising images rendered by MC path tracing. The input image data may include pixel color and its variance, as well as a set of auxiliary buffers that encode scene information (e.g., surface normal, albedo, depth, and their corresponding variances). In some embodiments, a CNN directly predicts the final denoised pixel value as a highly non-linear combination of the input features. In some other embodiments, a kernel-prediction neural network uses a CNN to estimate the local weighting kernels, which are used to compute each denoised pixel from its neighbors. In some embodiments, the input image can be decomposed into diffuse and specular components. The diffuse and specular components are then independently preprocessed, filtered, and postprocessed, before recombining them to obtain a final denoised image.

Abstract: A method, computer readable medium, and system are disclosed for sequential multi-tasking to generate coordinates of landmarks within images. The landmark locations may be identified on an image of a human face and used for emotion recognition, face identity verification, eye gaze tracking, pose estimation, etc. A neural network model processes input image data to generate pixel-level likelihood estimates for landmarks in the input image data and a soft-argmax function computes predicted coordinates of each landmark based on the pixel-level likelihood estimates.

Abstract: A facial expression recognition training system includes a training module, feature database, a capturing module, a recognition module and an adjusting module. The training module trains a facial expression feature capturing model according to known face images. The feature database stores known facial expression features of the known face images. The capturing module continuously captures first face images, and the facial expression feature capturing model outputs facial expression features of the first face images according to the first face images. The recognition module compares the facial expression features and the known facial expression features, and fit the facial expression features to the first known facial expression features that is one kind of the known facial expression feature accordingly. The adjusting module adjusts the facial expression feature capturing model to reduce the differences between the facial expression features and the known facial expression features.

Abstract: Disclosed embodiments relate to a character detection method and apparatus. In some embodiments, the method includes: using an image including an annotated word as an input to a machine learning model; selecting, based on a predicted result of characters inside an annotation region of the annotated word predicted and annotation information of the annotated word, characters for training the machine learning model from the characters inside the annotation region of the annotated word predicted; and training the machine learning model based on features of the selected characters. This implementation manner implements the full training of a machine learning model by using existing word level annotated images, to obtain a machine learning model capable of detecting characters in images, thereby reducing the costs for the training of a machine learning model capable of detecting characters in images.

Abstract: A mug shot acquisition system is disclosed that includes an image acquisition interface that is operative to receive digital mug shot images from an imaging device. An image viewing display is responsive to the image acquisition interface and operative to display the received digital mug shot images. One or more standards-based image adjustment software tools allow the digital mug shot images to be adjusted to meet at least one predetermined mug shot image uniformity standard. The system also includes a software interface responsive to requests to initiate one or more operations by the mug shot acquisition system, and an image export interface operative to export digital mug shot images adjusted based on the standards-based image adjustment software.

Abstract: A device can include a memory storing user payment data and another memory device storing instructions that cause the device to establish a communication between a separate device and the system based on a gesture associated with the system and via a wireless link between the system and the separate device, the communication being associated with a potential purchase, present, on the display, an instruction associated with the potential purchase, receive a single-interaction from the user of the system to confirm a payment for the potential purchase, the single-interaction including a security measure to prevent unauthorized purchases, retrieve, based on the single-interaction from the user, the user payment data from the memory and transmit the user payment data via the wireless link to the separate device to make a purchase.

Abstract: Aspects of the present disclosure relate to machine learning techniques for training a model to identify each of a number of different classes in images, based on training data where each training image may not be labeled in a complete manner with respect to the classes. The disclosed training techniques use a new label value to indicate when a ground truth value is unknown for a particular class, and do not penalize the machine learning network for output predictions that do not match the label value representing unknown ground truth. Some implementations of the training process can be regularized to impose sparsity on predicted classes in order to avoid false positive predictions.

Abstract: In a computer-implemented method and associated tangible non-transitory computer-readable medium, an image of a damaged vehicle may be analyzed to generate a repair estimate. A dataset populated with digital images of damaged vehicles and associated claim data may be used to train a deep learning neural network to learn damaged vehicle image characteristics that are predictive of claim data characteristics, and a predictive similarity model may be generated. Using the predictive similarity model, one or more similarity scores may be generated for a digital image of a newly damaged vehicle, indicating its similarity to one or more digital images of damaged vehicles with known damage level, repair time, and/or repair cost. A repair estimate may be generated for the newly damaged vehicle based on the claim data associated with images that are most similar to the image of the newly damaged vehicle.

Abstract: A system for estimating depth using a monocular camera may include one or more processors, a monocular camera, and a memory device. The monocular camera and the memory device may be operably connected to the one or more processors. The memory device may include an image capture, an encoder-decoder module, a semantic information generating module, and a depth map generating module. The modules may configure the one or more processors to executed by one or more processors cause the one or more processors to obtain a captured image from the monocular camera, generate a synthesized image based on the captured image wherein the style transfer module was trained using a generative adversarial network, generate, a feature map based on the synthesized image, generate semantic information based on the feature map, and generate a depth map based on the feature map and the semantic information.

Abstract: A convolutional neural network (CNN) processing method and apparatus. The apparatus may select, based on at least one of a characteristic of at least one kernel of a convolution layer or a characteristic of an input of the convolution layer, one operation mode from a first operation mode reusing a kernel, of the at least one kernel, and a second operation mode reusing the input, and perform a convolution operation based on the selected operation mode.

Abstract: An information processing device includes a storage unit that stores learning information acquired by machine learning, an input unit that acquires identification information, and a processing unit that performs recognition processing using the learning information that is specified by the storage unit on a basis of the identification information. An information processing method is executed by a processor, the information processing method includes storing learning information acquired by machine learning, acquiring identification information, and performing recognition processing using the learning information that is specified from storage on a basis of the identification information.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for processing images using recurrent attention. One of the methods includes determining a location in the first image; extracting a glimpse from the first image using the location; generating a glimpse representation of the extracted glimpse; processing the glimpse representation using a recurrent neural network to update a current internal state of the recurrent neural network to generate a new internal state; processing the new internal state to select a location in a next image in the image sequence after the first image; and processing the new internal state to select an action from a predetermined set of possible actions.

Abstract: Technologies for using a shifted neural network include a compute device to determine a shift-based activation function of the shifted neural network. The shift-based activation function is a piecewise linear approximation of a transcendental activation function and is defined by a plurality of line segments such that a corresponding slope of each line segment is a power of two. The compute device further trains the shifted neural network based on shift-based input weights and the determined shift-based activation function.

Abstract: The present disclosure provides an image capture, curation, and editing system that includes a resource-efficient mobile image capture device that continuously captures images. In particular, the present disclosure provides low power frameworks for controlling image sensor mode in a mobile image capture device. On example low power frame work includes a scene analyzer that analyzes a scene depicted by a first image and, based at least in part on such analysis, causes an image sensor control signal to be provided to an image sensor to adjust at least one of the frame rate and the resolution of the image sensor.