公开(公告)号：US20160155014A1

公开(公告)日：2016-06-02

申请号：US14906218

申请日：2014-07-23

Applicant: TELECOM ITALIA S.P.A.

Inventor： Massimo BALESTRI ,  Gianluca FRANCINI ,  Skjalg LEPSOY

IPC: G06K9/46

CPC classification number: G06K9/4671 ,  G06K9/4609 ,  G06K9/4652

Abstract: A method for identifying keypoints in a digital image including a set of pixels. Each pixel has associated thereto a respective value of an image representative parameter. The method includes approximating a filtered image. The filtered image depends on a filtering parameter and includes for each pixel of the image a filtering function that depends on the filtering parameter to calculate a filtered value of the value of the representative parameter of the pixel. The approximating includes: a) generating a set of base filtered images; each base filtered image is the image filtered with a respective value of the filtering parameter; b) for each pixel of at least a subset of the set of pixels, approximating the filtering function by a respective approximation function based on the base filtered images; the approximation function is a function of the filtering parameter within a predefined range of the filtering parameter.

Abstract translation: 一种用于识别包括一组像素的数字图像中的关键点的方法。 每个像素与其相关联地具有图像代表参数的相应值。 该方法包括近似滤波图像。 滤波后的图像取决于滤波参数，并且对于图像的每个像素包括依赖于滤波参数的滤波函数，以计算像素的代表性参数的值的滤波值。 近似值包括：a）生成一组基本滤波图像; 每个基本过滤图像是用过滤参数的相应值过滤的图像; b）对于所述像素集合的至少一个子集的每个像素，通过基于所述基本滤波图像的相应近似函数近似所述滤波函数; 近似函数是过滤参数的预定义范围内的过滤参数的函数。

公开(公告)号：US20200042871A1

公开(公告)日：2020-02-06

申请号：US16081693

申请日：2016-03-11

Applicant: TELECOM ITALIA S.p.A.

Inventor： Gianluca FRANCINI ,  Skjalg LEPSOY ,  Pedro PORTO BUARQUE DE GUSMAO

IPC: G06N3/08 ,  G06K9/62 ,  G06N3/04

Abstract: A method in which a convolutional neural network is configured to receive an input data structure including a group of values corresponding to signal samples and to generate a corresponding classification output indicative of a selected one among plural predefined classes. The convolutional neural network includes an ordered sequence of layers, each configured to receive a corresponding layer input data structure including a group of input values, and generate a corresponding layer output data structure including a group of output values by convolving the layer input data structure with at least one corresponding filter including a corresponding group of weights. The layer input data structure of the first layer of the sequence corresponds to the input data structure. The layer input data structure of a generic layer of the sequence different from the first layer corresponds to the layer output data structure generated by a previous layer in the sequence.

公开(公告)号：US20210142175A1

公开(公告)日：2021-05-13

申请号：US17251508

申请日：2019-07-18

Applicant: TELECOM ITALIA S.p.A.

Inventor： Attilio FIANDROTTI ,  Gianluca FRANCINI ,  Skjalg LEPSOY ,  Enzo TARTAGLIONE

IPC: G06N3/08 ,  G06N3/04 ,  G05D1/00

Abstract: A method includes providing a neural network having a set of weights. The neural network receives an input data structure for generating a corresponding output array according to values of the set of weights. The neural network is trained to obtain a trained neural network. The training includes setting values of the set of weights with a gradient descent algorithm which exploits a cost function including a loss term and a regularization term. The trained neural network is deployed on a device through a communication network, and used by the device. The regularization term is based on a rate of change of elements of the output array caused by variations of the set of weights values.