公开(公告)号：US20240233335A1

公开(公告)日：2024-07-11

申请号：US18616599

申请日：2024-03-26

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Dequan YU ,  Yin ZHAO ,  Haitao YANG

IPC: G06V10/77 ,  G06V10/774 ,  G06V10/82

CPC classification number: G06V10/7715 ,  G06V10/774 ,  G06V10/82

Abstract: A feature map processing method includes: determining P target strides based on a preset correspondence between a stride and a feature map size range and a size of a target feature map, where P is a positive integer; and invoking a neural network model to process the target feature map, to obtain a processing result of the target feature map, where the neural network model includes P dynamic stride modules, a stride of a dynamic stride module of the P dynamic stride modules is a target stride corresponding to a dynamic stride module in the P target strides.

公开(公告)号：US20240046055A1

公开(公告)日：2024-02-08

申请号：US18264551

申请日：2022-01-21

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Haoren WANG ,  Dequan YU ,  Yi REN ,  Haosheng GAO ,  Chunliang LIU ,  Ping SONG

IPC: G06K7/14

CPC classification number: G06K7/1417

Abstract: Methods and apparatuses for recognizing a two-dimensional code are disclosed. In an implementation, a method comprises: identifying, by an electronic device, a two-dimensional code, wherein the two-dimensional code comprises an image region in the center and a ring region surrounding the image region, wherein the ring region comprises a code region including a first code region and a second code region, a first spacing region, and a second spacing region, wherein the first spacing region and the second spacing region are arranged between the first code region and the second code region, determining values corresponding to a plurality of code elements in the code region, and recognizing, by the electronic device based on the values, first information corresponding to the two-dimensional code.

公开(公告)号：US20250156712A1

公开(公告)日：2025-05-15

申请号：US19019814

申请日：2025-01-14

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Dequan YU ,  Yin ZHAO

IPC: G06N3/08 ,  G06N3/04

Abstract: This application discloses a model training method, including: obtaining training data; using the training data as an input of a model, and in a training process of the model, calculating a parameter by using a first precision range, to obtain a calculated value; and if the calculated value overflows the first precision range, recalculating the parameter by using a second precision range, and performing iterative training on the model for one or more times by using a recalculated parameter, where the second precision range includes the first precision range, or the second precision range partially overlaps the first precision range.

公开(公告)号：US20250142079A1

公开(公告)日：2025-05-01

申请号：US19010423

申请日：2025-01-06

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Jue MAO ,  Dequan YU ,  Sheng QIAN ,  Yin ZHAO

IPC: H04N19/136 ,  H04N19/167 ,  H04N19/17 ,  H04N19/463

Abstract: Embodiments of this application provide an encoding and a device. The encoding method includes: obtaining an image; generating feature maps of C channels based on the image, the feature maps including feature values of feature points; generating estimated information matrices of the C channels based on the feature maps of the C channels; grouping the C channels into N channel groups; for at least one target channel group in the N channel groups, determining, based on at least one feature value of at least one encoded feature point corresponding to the target channel group and an estimated information matrix corresponding to the target channel group, at least one probability distribution parameter of a to-be-encoded feature point corresponding to the target channel group; determining, based on the at least one probability distribution parameter, probability distribution of the to-be-encoded feature point; and encoding the to-be-encoded feature point based on the probability distribution.