Abstract: An approach is provided in which a knowledge manager selects an extraction layer from a convolutional neural network that was trained on an initial set of images. The knowledge manager processes subsequent images obtained from crawling a computer network that includes extracting image feature sets of the subsequent images from the selected extraction layer and generating tags from metadata associated with the subsequent images. In turn, the knowledge manager receives a new image, extracts a new image feature set from the selected extraction layer, and assigns one or more of the tags to the new image based upon evaluating the new image feature set to the image features sets of the subsequent images.

Abstract: Techniques are disclosed for sorting an input data set. A sort tool determines a distribution of values of a data set that includes a plurality of data records. The sort tool partitions the data set into a plurality of subsets based on the distribution. Each of the data records is inserted into one of the subsets based on a corresponding sort value of the data record. The sort tool identifies one or more of the subsets that contain at least two distinct sort values. In each of the identified subsets, the data records are sorted by a corresponding sort value of the data record.

Abstract: Techniques are disclosed for sorting an input data set. A sort tool determines a distribution of values of a data set that includes a plurality of data records. The sort tool partitions the data set into a plurality of subsets based on the distribution. Each of the data records is inserted into one of the subsets based on a corresponding sort value of the data record. The sort tool identifies one or more of the subsets that contain at least two distinct sort values. In each of the identified subsets, the data records are sorted by a corresponding sort value of the data record.

Abstract: In an example embodiment, a deep convolutional neural network (DCNN) is created to assign a professionalism score to an input image. The professionalism score indicates a perceived professionalism of a subject of the input image. The DCNN is designed to automatically learn features of images relevant to the professionalism through a training process.

Abstract: A method and apparatus for cross-domain medical image synthesis is disclosed. A source domain medical image is received. A synthesized target domain medical image is generated using a trained contextual deep network (CtDN) to predict intensities of voxels of the target domain medical image based on intensities and contextual information of voxels in the source domain medical image. The contextual deep network is a multi-layer network in which hidden nodes of at least one layer of the contextual deep network are modeled as products of intensity responses and contextual response.

Abstract: An image processing system and/or method obtains source images in which a damaged vehicle is represented, and performs image processing techniques to determine, predict, estimate, and/or detect damage that has occurred at various locations on the vehicle. The image processing techniques may include generating a composite image of the damaged vehicle, aligning and/or isolating the image, applying convolutional neural network techniques to the image to generate damage parameter values, where each value corresponds to damage in a particular location of vehicle, and/or other techniques. Based on the damage values, the image processing system/method generates and displays a heat map for the vehicle, where each color and/or color gradation corresponds to respective damage at a respective location on the vehicle. The heat map may be manipulatable by the user, and may include user controls for displaying additional information corresponding to the damage at a particular location on the vehicle.

Abstract: According to a method and apparatus disclosed herein, an electronic register advantageously logs raw sensor data without converting the raw sensor data into physical-domain measurements, and without need for being configured to understand or process such data, or account or compensate for any sensor installation particulars. Instead, the electronic register advantageously stores conversion data in association with each sensor interface circuit being used to collect raw sensor data from a corresponding external sensor, and it provides the conversion data to an external device, in association with read-out of the raw sensor data logged by the electronic unit. The conversion data provides the mathematical expression, along with any compensation or adjustment values needed, to convert the raw sensor data into corresponding physical-domain measurements.

Abstract: Briefly, embodiments of methods and/or systems of detecting and image of a human face in a digital image are disclosed. For one embodiment, as an example, parameters of a neural network may be developed to generate object labels for digital images. The developed parameters may be refined by a neural network to generate signal sample value levels corresponding to probability that a human face may be depicted at a localized region of a digital image.

Abstract: A method upsamples an image using a non-linear fully connected neural network to produce only global details of an upsampled image and interpolates the image to produce a smooth upsampled image. The method concatenates the global details and the smooth upsampled image into a tensor and applies a sequence of nonlinear convolutions to the tensor using a convolutional neural network to produce the upsampled image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.

Abstract: An information processing apparatus comprises: a programmable circuit unit comprising a partial reconfiguration unit; a storage unit used by each of logic circuits configured in the partial reconfiguration unit; and a control unit that controls a logic circuit that becomes an access destination, in accordance with receiving an access command, wherein the control unit compares an address space indicating the access destination of the access command with the signal that is output from the partial reconfiguration unit due to the partial reconfiguration unit being configured using circuit information included in the configuration data, and controls to set as an access destination the logic circuit configured in the partial reconfiguration unit outputting the signal matching the address space indicating the access destination of the access command.

Abstract: A system and method for hybrid machine translation approach is based on a statistical transfer approach using statistical and linguistic features. The system and method may be used to translate from one language into another. The system may include at least one database, a rule based translation module, a statistical translation module and a hybrid machine translation engine. The database(s) store source and target text and rule based language models and statistical language models. The rule based translation module translates source text based on the rule based language models. The statistical translation module translates source text based on the statistical language models. A hybrid machine translation engine, having a maximum entropy algorithm, is coupled to the rule based translation module and the statistical translation module and is capable of translating source text into target text based on the rule based and statistical language models.

Abstract: Systems and methods are disclosed for segregating target individuals represented in a probe digital image from background pixels in the probe digital image. In particular, in one or more embodiments, the disclosed systems and methods train a neural network based on two or more of training position channels, training shape input channels, training color channels, or training object data. Moreover, in one or more embodiments, the disclosed systems and methods utilize the trained neural network to select a target individual in a probe digital image. Specifically, in one or more embodiments, the disclosed systems and methods generate position channels, training shape input channels, and color channels corresponding the probe digital image, and utilize the generated channels in conjunction with the trained neural network to select the target individual.

Abstract: Techniques for enhancing an acoustic echo canceller based on visual cues are described herein. The techniques include changing adaptation of a filter of the acoustic echo canceller, calibrating the filter, or reducing background noise from an audio signal processed by the acoustic echo canceller. The changing, calibrating, and reducing are responsive to visual cues that describe acoustic characteristics of a location of a device that includes the acoustic echo canceller. Such visual cues may indicate that no human being is present at the location, that some subject(s) are engaged in speaking or sound generating activities, or that motion associated with an echo path change has occurred at the location.

Abstract: Known methods for generating super-resolved images from single input images have various disadvantages. An improved method for generating a super-resolved image from a single low-resolution input image comprises up-scaling the input image to generate an initial version of the super-resolved image, searching, for each patch of the low-resolution input image, similar low-resolution patches in first search windows within down-sampled versions of the input image, and determining, in less down-sampled versions of the input image, high-resolution patches that correspond to the similar low-resolution patches. The determined high-resolution patches are cropped, a second search window is determined in the initial version of the super-resolved image, and a best-matching position for each cropped high-resolution patch is searched within the second search window. Finally, each cropped high-resolution patch is added to the super-resolved image at its respective best-matching position.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for computing a layer output for a convolutional neural network layer, the method comprising: receiving a plurality of activation inputs; forming a plurality of vector inputs from the plurality of activation inputs, each vector input comprising values from a distinct region within the multi-dimensional matrix; sending the plurality of vector inputs to one or more cells along a first dimension of the systolic array; generating a plurality of rotated kernel structures from each of the plurality of kernel; sending each kernel structure and each rotated kernel structure to one or more cells along a second dimension of the systolic array; causing the systolic array to generate an accumulated output based on the plurality of value inputs and the plurality of kernels; and generating the layer output from the accumulated output.

Abstract: A digital camera includes an image optimization engine configured to generate an optimized image based on a raw image captured by the digital camera. The image optimization engine implements one or more machine learning engines in order to select rendering algorithms and rendering algorithm arguments that may then be used to render the raw image.

Abstract: A method and system for converting low-dose mammographic images with much noise into higher quality, less noise, higher-dose-like mammographic images, using of a trainable nonlinear regression (TNR) model with a patch-input-pixel-output scheme, which can be called a call pixel-based TNR (PTNR). An image patch is extracted from an input mammogram acquired at a reduced x-ray radiation dose (lower-dose), and pixel values in the patch are entered into the PTNR as input. The output of the PTNR is a single pixel that corresponds to a center pixel of the input image patch. The PTNR is trained with matched pairs of mammograms, inputting low-dose mammograms together with corresponding desired standard x-ray radiation dose mammograms (higher-dose), which are ideal images for the output images. Through the training, the PTNR learns to convert low-dose mammograms to high-dose-like mammograms. Once trained, the trained PTNR does not require the higher-dose mammograms anymore.

Abstract: A first set of contextual dimensions is generated from one or more textual descriptions associated with a given event, which includes one or more examples. A second set of contextual dimensions is generated from one or more visual features associated with the given event, which includes one or more visual example recordings. A similarity structure is constructed from the first set of contextual dimensions and the second set of contextual dimensions. One or more of the textual descriptions is matched with one or more of the visual features based on the similarity structure.

Abstract: For digital pathology imaging, intelligent processing, such as automatic recognition or content-based retrieval, is one significant benefit that drives the wide application of this technology. Before any intelligent processing on pathology images, every image is converted into a feature vector which quantitatively capture its visual characteristics. An algorithm characterizing pathology images with statistical analysis of local responses of neural networks is described herein. The algorithm framework enables extracting sophisticated textural features that are well adapted to the image data of interest.

Abstract: Aspects described herein are directed towards methods, computing devices, systems, and computer-readable media that apply scattering operations to extracted visual features of audiovisual input to generate predictions regarding the speech status of a subject. Visual scattering coefficients generated according to one or more aspects described herein may be used as input to a neural network operative to generate the predictions regarding the speech status of the subject. Predictions generated based on the visual features may be combined with predictions based on audio input associated with the visual features. In some embodiments, the extracted visual features may be combined with the audio input to generate a combined feature vector for use in generating predictions.

Abstract: An identification apparatus performs classification using a plurality of classifiers, and calculates the reliability of its classification result. A data obtaining unit obtains input data. A feature quantity obtaining unit obtains a feature quantity corresponding to the input data. A plurality of classifiers receive input of the feature quantity and perform classification based on the input feature quantity. An identification unit inputs the feature quantity into each of the classifiers, and generates a single second classification result based on a plurality of classification results obtained from the classifiers. A reliability generation unit generates a reliability of the second classification result based on variations across the plurality of classification results.

Abstract: Systems and techniques are provided for pruning a node from a possible nodes list for Hidden Markov Model with label transition node pruning. The node may be a label transition node. A frame may be at a predicted segmentation point in decoding input with the Hidden Markov Model. The node may be scored at the frame. The node may be pruned from the possible nodes list for the frame when score for the node is greater than the sum of a best score among nodes on the possible nodes list for the frame and a beam threshold minus a penalty term. A possible nodes list may be generated for a subsequent frame using label selection. A second node may be pruned from the possible nodes list for the subsequent frame with early pruning.

Abstract: Systems, methods and computer-accessible mediums for modifying an image(s) can be provided. For example, first image information for the image(s) can be received. Second image information can be generated by separating the first image information into at least two overlapping images. The image(s) can be modified using a prediction procedure based on the second image information.

Abstract: A system, method, and computer program product for assigning an aesthetic score to an image. A method of the present invention includes receiving an image. The method further includes executing a neural network on the image to generate learned features. The method further includes applying a machine-learned model to assign an aesthetic score to the image, where a more aesthetically-pleasing image is given a higher aesthetic score and a less aesthetically-pleasing image is given a lower aesthetic score. The learned features are inputs to the machine-learned model.

Abstract: The present disclosure is to generate a high-quality image by correcting a predetermined correction target image based on a plurality of input images. In an image processing device 3, an image correcting section 160 detects a user tap gesture on a touch panel 250. When the position of the tap gesture is within foreground candidate areas detected by a foreground candidate area detecting section 140, the image correcting section 160 corrects a base image set by a base image setting section 120 in the areas corresponding to the foreground candidate areas.

Abstract: Embodiments of a computer-implemented method for training a convolutional neural network (CNN) that is pre-trained using a set of color images are disclosed. The method comprises receiving a training dataset including multiple multidimensional images, each multidimensional image including a color image and a depth image; performing a fine-tuning of the pre-trained CNN using the depth image for each of the plurality of multidimensional images; obtaining a depth CNN based on the pre-trained CNN, wherein the depth CNN is associated with a first set of parameters; replicating the depth CNN to obtain a duplicate depth CNN being initialized with the first set of parameters; and obtaining a depth-enhanced color CNN based on the duplicate depth CNN being fine-tuned using the color image for each of the plurality of multidimensional images, wherein the depth-enhanced color CNN is associated with a second set of parameters.

Abstract: The present disclosure provides a method that includes capturing a first image of a mask in a first exposure apparatus using a first exposure source and a first imaging sensor; capturing a second image of the mask in a second exposure apparatus using a second exposure source and a second imaging sensor; comparing the first image of the mask and the second image of the mask for a difference therebetween; and determining an action according to the difference.

Abstract: A digital image having a plurality of pixels is analyzed to detect a fire condition. A first color parameter is determined from image color values of pixels of the image. A plurality of fuzzy membership functions correlated to image colors are defined, the plurality of fuzzy membership functions including a first fuzzy color membership function having a trend defined by said first color parameter. A fuzzy inference procedure is applied to pixels of the image to determine whether a fire condition is indicated by the digital image.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for identifying objects in images. One of the methods includes receiving an input image; down-sampling the input image to generate a second image; generating a respective first score for each of the plurality of object categories; selecting an initial patch of the input image; generating a respective second score for each of the plurality of object categories; and generating a respective third score for each of the plurality of object categories from the first scores and the second scores, wherein the respective third score for each of the plurality of object categories represents a likelihood that the input image contains an image of an object belonging to the object category.

Abstract: A system, apparatus, method, and computer-readable medium for optimizing classifiers are disclosed. The optimization process can include receiving one or more training examples. The optimization process can further include assigning a loss parameter to each training example. The optimization process can further include optimizing each loss parameter of each training sample based on a sample variance of each training example using a non-linear function. The optimization process can further include estimating a classifier from the one or more weighted training samples. The optimization process can further include assigning a loss parameter to the classifier based on a number of training examples that the classifier correctly classified and a number of training examples that the classifier incorrectly classified. The optimization process can further include adding the weighted classifier to an overall classifier.

Abstract: A method, apparatus and computer program product are provided in order to manage a medical imaging study display environment. The method includes receiving, from at least one display device coupled to a medical imaging workstation, display environment context data. The method also includes extracting, by processing circuitry, metadata from a medical imaging study accessed by the medical imaging workstation, determining based at least in part on the metadata, image viewing context data, determining, based on the display environment context data and the image viewing context data, whether a display environment of the medical imaging workstation complies with one or more rules for viewing the medical imaging study, and in response to determining that the display environment of the medical imaging workstation does not comply with the one or more rules, performing at least one action before allowing viewing of the medical imaging study to proceed.

Abstract: A convolutional neural network is trained to analyze input data in various different manners. The convolutional neural network includes multiple layers, one of which is a convolution layer that performs a convolution, for each of one or more filters in the convolution layer, of the filter over the input data. The convolution includes generation of an inner product based on the filter and the input data. Both the filter of the convolution layer and the input data are binarized, allowing the inner product to be computed using particular operations that are typically faster than multiplication of floating point values. The possible results for the convolution layer can optionally be pre-computed and stored in a look-up table. Thus, during operation of the convolutional neural network, rather than performing the convolution on the input data, the pre-computed result can be obtained from the look-up table.

Abstract: An image processing device includes an input reception section that receives a learning image and a correct answer label, a processing section that performs a process that generates classifier data and a processing target image, and a storage section. The processing section generates the processing target image that is the entirety or part of the learning image, calculates a feature quantity of the processing target image, generates the classifier data based on training data that is a set of the feature quantity and the correct answer label assigned to the learning image that corresponds to the feature quantity, generates an image group based on the learning image or the processing target image, classifies each image of the image group using the classifier data to calculate a classification score of each image, and regenerates the processing target image based on the classification score and the image group.

Abstract: Systems and methods for detecting a visual characteristic of interest within an image are disclosed. An example method involves obtaining an image that includes at least one pixel representing a visual characteristic of interest, creating a first sequence and a second sequence of bitwise data from values associated with the pixel, and converting these bitwise sequences into a first sequence of integers and a second sequence of integers. Using a distance function, a similarity metric is determined between the first sequence of integers and the second sequence of integers. Based on the similarity metric, a third sequence of integers is created and stored. The third sequence of integers can be used to facilitate the identification of the visual characteristic of interest in other images.

Abstract: A local feature descriptor for a point in an image is generated over multiple levels of an image scale space. The image is gradually smoothened to obtain a plurality of scale spaces. A point may be identified as the point of interest within a first scale space from the plurality of scale spaces. A plurality of image derivatives is obtained for each of the plurality of scale spaces. A plurality of orientation maps is obtained (from the plurality of image derivatives) for each scale space in the plurality of scale spaces. Each of the plurality of orientation maps is then smoothened (e.g., convolved) to obtain a corresponding plurality of smoothed orientation maps. Therefore, a local feature descriptor for the point may be generated by sparsely sampling a plurality of smoothed orientation maps corresponding to two or more scale spaces from the plurality of scale spaces.

Abstract: Image processing circuitry may be used to detect imaging errors of a pixel array, including imaging errors that would otherwise be difficult to detect visually. The image processing circuitry may identify a subset of the pixels that are characterized by burst noise such as RTS noise. The image processing circuitry may monitor the identified subset of the pixels to detect imaging errors. The circuitry may maintain histograms of image values for the subset of pixels from previously captured images. The image processing circuitry may update the histograms using each new captured image and determine whether the updated histograms have tri-modal distributions. In response to determining that the updated histograms do not have tri-modal distributions of values, the circuitry may identify that an imaging error has occurred. If desired, expected differences in dark current across multiple exposure settings may be used to help identify imaging errors.

Abstract: A method for estimating illumination of an image captured by a digital system is provided that includes computing a feature vector for the image, identifying at least one best reference illumination class for the image from a plurality of predetermined reference illumination classes using the feature vector, an illumination classifier, and predetermined classification parameters corresponding to each reference illumination class, and computing information for further processing of the image based on the at least one best reference illumination class, wherein the information is at least one selected from a group consisting of color temperature and white balance gains.

Abstract: Systems and methods for notification and privacy management of online photos and videos are herein disclosed. According to one embodiment, a computer-implemented method includes detecting a first feature from a first image belonging to an image source. The first image includes a tag. The computer-implemented method further includes determining a first feature location of the first feature, determining a first tag location of the tag, extracting a first feature signature from the first feature based on a relationship between the first feature location and the first tag location, detecting a second feature from a second image belonging to the image source, extracting a second feature signature from the second feature, performing a first comparison between the first feature signature and the second feature signature, and deriving a first similarity score based on a result of the first comparison.

Abstract: A convolutional neural network (CNN) is trained for font recognition and font similarity learning. In a training phase, text images with font labels are synthesized by introducing variances to minimize the gap between the training images and real-world text images. Training images are generated and input into the CNN. The output is fed into an N-way softmax function dependent on the number of fonts the CNN is being trained on, producing a distribution of classified text images over N class labels. In a testing phase, each test image is normalized in height and squeezed in aspect ratio resulting in a plurality of test patches. The CNN averages the probabilities of each test patch belonging to a set of fonts to obtain a classification. Feature representations may be extracted and utilized to define font similarity between fonts, which may be utilized in font suggestion, font browsing, or font recognition applications.

Abstract: In an embodiment, a processor includes a compression domain threshold filter coupled to a plurality of cores. The compression domain threshold filter is to: receive a sample vector of compressed data to be filtered; calculate, based at least on a first subset of the elements of the sample vector, an estimated upper bound value of a dot product of the sample vector and a steering vector; determine whether the estimated upper bound value of the dot product satisfies a filter threshold value; and in response to a determination that the estimated upper bound value of the dot product does not satisfy the filter threshold value, discard the sample vector without completion of a calculation of the dot product of the sample vector and the steering vector. Other embodiments are described and claimed.

Abstract: Methods and systems for bootstrapping an OCR engine for license plate recognition. One or more OCR engines can be trained utilizing purely synthetically generated characters. A subset of classifiers, which require augmentation with real examples, along how many real examples are required for each, can be identified. The OCR engine can then be deployed to the field with constraints on automation based on this analysis to operate in a “bootstrapping” period wherein some characters are automatically recognized while others are sent for human review. The previously determined number of real examples required for augmenting the subset of classifiers can be collected. Each subset of identified classifiers can then be retrained as the number of real examples required becomes available.

Abstract: Systems and methods for sequence transcription with neural networks are provided. More particularly, a neural network can be implemented to map a plurality of training images received by the neural network into a probabilistic model of sequences comprising P(S|X) by maximizing log P(S|X) on the plurality of training images. X represents an input image and S represents an output sequence of characters for the input image. The trained neural network can process a received image containing characters associated with building numbers. The trained neural network can generate a predicted sequence of characters by processing the received image.

Abstract: To provide an information processing device for outputting estimated nutrition value information on at least one dish or a drink through simplified processing. The information processing device holds as reference information a captured image of at least one dish or a drink and nutrition value information set for the item of food or drink included in the image so as to be correlated to each other, receives an image including at least one dish or a drink, and retrieves a plurality of images similar to the image received from among the reference information held. Further, the information processing device calculates a statistic of a plurality of nutrition value information pieces correlated to the plurality of respective images retrieved, and outputs the statistic calculated.

Abstract: A method re-identifies objects in a pair of images by applying a convolutional neural network (CNN). Each layer in the network operates on an output of a previous layer. The layers include a first convolutional layer and a first max pooling layer to determine a feature map, a cross-input neighborhood differences layer to produce neighborhood difference maps, a patch summary layer to produce patch summary feature maps, a first fully connected layer to produce a feature vector representing higher order relationships in the patch summary feature maps, a second fully connected layer to produce two scores representing positive pair and negative pair classes, and a softmax layer to produce positive pair and negative pair probabilities. Then, the positive pair probability is output to signal whether the two images represent the same object or not.

Abstract: A system and method for facilitating patent grant and patent application claims examination; including the functions of automated importing of patent claims, automated parsing of the claims into their hierarchy, and compression/expansion of the parsed claims to/from the independent claim level.

Abstract: Various embodiments describe systems and methods enable a computing device of a user to capture an image of a gift card, or other such monetary device containing a code, with a camera or otherwise receive an image of that gift card. The computing device can be configured to recognize codes, such as digit claim codes, of the gift card by using one or more image processing, computer vision, and/or machine learning algorithms. After a successful detection and verification of a claim code, money or funds deposited in, or otherwise available from, an account associated with the gift card can be utilized, such as applied to a purchase or deposited into the user's account. In many instances, a user interface (UI) can be provided on the computing device for the user to use to capture an image of a gift card and redeem the funds from the corresponding card.

Abstract: A method, non-transitory computer readable medium, and apparatus for classifying machine printed text and handwritten text in an input are disclosed. For example, the method defines a perspective for an auto-encoder, receives the input for the auto-encoder, wherein the input comprises a document comprising the machine printed text and the handwritten text, performs an encoding on the input using an auto-encoder to generate a classifier, applies the classifier on the input and generates an output that separates the machine printed text and the handwritten text in the input based on the classifier in accordance with the perspective.

Abstract: The invention relates to an image processing device (1) for finding corresponding first and second regions in two image data sets of an object. In a first image data set a source line and in a second image data set a corresponding target line are determined depending on reference regions detectable in both image data sets. A first region in the first image data set is projected onto the source line, thereby dividing the source line into two source sub-lines and defining a source ratio as the ratio of the length of one of the source sub-lines to the length of the entire source line. A second region in the second image data set is then determined such that a projection of the second region onto the target line leads to a corresponding target ratio which is similar to the source ratio.

Abstract: A method for sending a screen image by a computing device is described. A modification image representing a modified region of the screen image is determined. The modification image has a lossless format. Location information is encoded in one or more alpha channels of the modification image. The modification image is sent.