Nippon Telegraph and Telephone Corporation

Tetsuya Kinebuchi

Jun Shimamura

Yukito Watanabe

Takashi Hosono

Shuhei Tarashima

Shingo Ando

Hitoshi Niigaki

Yasuhiro Yao

Kazuhiko Murasaki

Ken Tsutsuguchi

Takayuki Umeda

Kaori Kumagai

Shunsuke Tsukatani

The disclosed technology describes determining and registering dictionary data to diagnose a deteriorating state of a surface of a diagnose object. The method comprises receiving spectral distribution information of deteriorating surface regions of a target object. Given the spectral distribution information and a predetermined spectral distribution information of a reference object, the present technology determines a reference reflectance value of the target object and registers the reference reflectance value of the target object as dictionary data. The reference reflectance value is approximately the same regardless of a progressing state of deterioration of a surface of the target object. Given the dictionary data, the present invention estimates a deterioration state of a surface of a diagnose object under a variety of type of light sources with accuracy, without measuring spectral distribution information about a light source used at the time of measuring spectral data of the diagnose object.

Diagnostic dictionary registering device, diagnosing device, method, program, and data structure

The present invention makes it possible to estimate, with high precision, a candidate region indicating each of multiple target objects included in an image. A parameter determination unitdetermines parameters to be used when detecting a boundary line of an imagebased on a ratio between a density of boundary lines included in an imageand a density of boundary lines in a region indicated by region informationindicating the region including at least one of the multiple target objects included in the image. A boundary line detection unitdetects the boundary line in the imageusing the parameter. For each of the multiple target objects included in the image, the region estimation unitestimates the candidate region of the target object based on the detected boundary line.

Candidate region estimation device, candidate region estimation method, and program

Simultaneous learning of a plurality of different tasks and domains, with low costs and high precision, is enabled. A learning unit, on the basis of learning data, uses a target encoder that takes data of a target domain as input and outputs a target feature expression, a source encoder that takes data of a source domain as input and outputs a source feature expression, a common encoder that takes data of the target domain or the source domain as input and outputs a common feature expression, a target decoder that takes output of the target encoder and the common encoder as input and outputs a result of executing a task with regard to data of the target domain, and a source decoder that takes output of the source encoder and the common encoder as input and outputs a result of executing a task with regard to data of the source domain, to learn so that the output of the target decoder matches training data, and the output of the source decoder matches training data.

Model learning device, model learning method, and program

It is possible to determine a geometric transformation matrix representing geometric transformation between an input image and a template image with high precision. A geometric transformation matrix/inlier estimation sectiondetermines a corresponding point group serving as inliers, and estimates the geometric transformation matrix representing the geometric transformation between the input image and the template image. A scatter degree estimation sectionestimates scatter degree of the corresponding points based on the corresponding point group serving as inliers. A plane tracking convergence determination threshold calculation sectioncalculates a threshold used in convergence determination when iterative update of the geometric transformation matrix in a plane tracking sectionis performed based on the estimated scatter degree. The plane tracking sectioniterates update of the geometric transformation matrix so as to minimize a difference in value to pixels of a geometric transformation image obtained by transforming one of the input image and the templated image using the geometric transformation matrix and corresponding pixels until it is determined that convergence has been completed using the calculated threshold.

Template orientation estimation device, method, and program

Even if an object to be detected is not remarkable in images, and the input includes images including regions that are not the object to be detected and have a common appearance on the images, a region indicating the object to be detected is accurately detected. A local feature extraction unitextracts a local feature of a feature point from each image included in an input image set. An image-pair common pattern extraction unitextracts, from each image pair selected from images included in the image set, a common pattern constituted by a set of feature point pairs that have similar local features extracted by the local feature extraction unitin images constituting the image pair, the set of feature point pairs being geometrically similar to each other. A region detection unitdetects, as a region indicating an object to be detected in each image included in the image set, a region that is based on a common pattern that is omnipresent in the image set, of common patterns extracted by the image-pair common pattern extraction unit

Object detection device, method, and program

Objectness indicating a degree of accuracy of a single object is accurately estimated. An edge detection unitdetects an edge for a depth image, an edge density/uniformity calculation unitcalculates an edge density on the periphery of a candidate region, an edge density inside the candidate region, and edge uniformity on the periphery of the candidate region. An objectness calculation unitcalculates the objectness of the candidate region based on the edge density on the periphery of the candidate region, the edge density inside the candidate region, and the edge uniformity on the periphery of the candidate region.

Object likelihood estimation device, method, and program

The present invention relates to representing image features used by a convolutional neural network (CNN) to identify concepts in an input image. The CNN includes a plurality of filters in each of a plurality of layers. The method generates the CNN based on a set of images for training with predetermined concepts in regions of the set of images. For a select layer of the CNN, the method generates integrated maps, Each integrated map is based on a set of feature maps in a cluster and relevance between the set of feature maps for the select layer and a region representing one of the features in the image data. The method provides a pair of a feature representation visualization image of a feature in the select layer and a concept information associated with the integration map.

Feature representation device, feature representation method, and program

A large-scale point cloud having no limitation on the range or the number of points is set as an object, and labels are attached to the points constituting the object regardless of the type of object. A three-dimensional point cloud label learning apparatusA includes a ground-based height calculation unit that receives a three-dimensional point cloud and a ground surface height as inputs and outputs a three-dimensional point cloud with a ground-based height, an intensity-RGB conversion unit that receives the three-dimensional point cloud with the ground-based height as an input and outputs an intensity-RGB converted three-dimensional point cloud with the ground-based height, a supervoxel clustering unit that receives the intensity-RGB converted three-dimensional point cloud with the ground-based height, a point cloud label for learning, and a clustering hyperparameter as inputs and outputs supervoxel data with a correct answer label, and a deep neural network learning unit that receives the supervoxel data with the correct answer label and a deep neural network hyperparameter as inputs and outputs a learned deep neural network parameter.

Three-dimensional point cloud label learning device, three-dimensional point cloud label estimating device, three-dimensional point cloud label learning method, three-dimensional point cloud label estimating method, and program

An object is to make it possible to precisely infer a geometric transformation matrix for transformation between an image and a reference image representing a plane region even if correspondence to the reference image cannot be obtained. A first line segment group extraction unitextracts, out of a line segment group in an image, line segments that correspond to a direction that is parallel or perpendicular to a side of a rectangle included in the image, from the inside of the rectangle, takes the extracted line segments to be a first line segment group, and extracts a plurality of line segments different from the first line segment group out of the line segment group. An endpoint detection unitdetects four intersection points between ends of the image and two line segments that are selected from line segments that correspond to a direction that is parallel or perpendicular to a side of the rectangle and are extracted from a plurality of line segments obtained by transforming the different line segments using an affine transformation matrix in which an angle of the first line segment group relative to a reference direction of the image is used as a rotation angle. A homography matrix inferring unitcomputes a geometric transformation matrix based on the affine transformation matrix and a homography matrix computed based on correspondence between the four intersection points and the four vertexes of the rectangle in a reference image.

Growing community of inventors

Tokyo, Japan

Tetsuya Kinebuchi