公开(公告)号：US20210081729A1

公开(公告)日：2021-03-18

申请号：US16984231

申请日：2020-08-04

Applicant: Beijing Baidu Netcom Science Technology Co., Ltd.

Inventor： Xiangkai HUANG ,  Leyi WANG ,  Lei NIE ,  Siyu AN ,  Minghao LIU ,  Jiangliang GUO

IPC: G06K9/72 ,  G06F40/30 ,  G06K9/20 ,  G06K9/00

Abstract: The present application discloses a method for image text recognition, an apparatus, a device, and a storage medium, and relates to image processing technologies in the field of cloud computing. A specific implementation is: acquiring an image to be processed, where at least one text line exists in the image to be processed; processing each text line in the image to be processed to obtain a composite encoded vector corresponding to each word in each text line, where the composite encoded vector carries semantic information and position information; and determining a text recognition result of the image to be processed according to the semantic information and the position information carried in the composite encoded vector corresponding to each word in each text line. This technical solution can accurately distinguish adjacent fields with small pixel spacing in the image and improve the accuracy of text recognition in the image.

公开(公告)号：US20210073973A1

公开(公告)日：2021-03-11

申请号：US16871633

申请日：2020-05-11

Applicant: BEIJING BAIDU NETCOM SCIENCE TECHNOLOGY CO., LTD.

Inventor： Jianfa ZOU ,  Ye SU ,  Minghao LIU ,  Lei NIE ,  Jiabing LENG ,  Yawei WEN ,  Tehui HUANG ,  Yulin XU ,  Jiangliang GUO ,  Xu LI

IPC: G06T7/00 ,  G06T7/80

Abstract: Provided are a method and an apparatus for component fault detection based on an image, and a specific implementation is: when it is determined that an image shot by an image pickup apparatus for a component to be tested with a first shooting parameter does not meet a preset condition, adjusting the first shooting parameter to a second shooting parameter; controlling the image pickup apparatus to shoot for the component to be tested with the second shooting parameter to obtain a first image that meets the preset condition; and performing fault detection on the component to be tested according to the first image. The image pickup apparatus can be adjusted in real time, so that the image can be used for fault detection only when meeting the preset condition, thereby the image is kept stable, and the accuracy rate of component fault identification based on an image is improved.

公开(公告)号：US20230252295A1

公开(公告)日：2023-08-10

申请号：US18300217

申请日：2023-04-13

Applicant: Beijing Baidu Netcom Science Technology Co., Ltd.

Inventor： Shuo LI ,  Lei NIE ,  Xiaowen XU

IPC: G06N3/08

CPC classification number: G06N3/08 ,  G06V10/778

Abstract: A method of generating a multimodal set of samples for an intelligent inspection, and a training method, which relate to a field of an artificial intelligence technology, in particular to fields of deep learning, natural language processing, speech technology, computer vision, big data and so on. The method of generating a multimodal set of samples includes: inputting an environmental sample in a collected multimodal set of environmental samples into a single-modal model matched with a modality of the environmental sample, so as to obtain a model processing result corresponding to the environmental sample; determining an initial set of samples from the multimodal set of environmental samples according to the model processing result; and processing the initial set of samples by means of an active learning, so as to determine the multimodal set of samples.

公开(公告)号：US20230214985A1

公开(公告)日：2023-07-06

申请号：US18181967

申请日：2023-03-10

Applicant: Beijing Baidu Netcom Science Technology Co., Ltd.

Inventor： Shuo LI ,  Feng HUANG ,  Lei NIE ,  Xuepeng ZHAO ,  Luyan CHEN

IPC: G06T7/00

CPC classification number: G06T7/0004 ,  G06T2207/20081

Abstract: A product recognition method and device, a model training method and device, and an electronic device are provided. The product recognition method includes: obtaining image data of a product; performing defect recognition on the image data based on a first recognition model, to obtain a first recognition result, wherein the first recognition model is configured to recognize a defective product; performing qualification recognition on the image data based on a second recognition model to obtain a second recognition result, wherein the second recognition model is configured to recognize a qualified product; determining a target recognition result of the product based on the first recognition result and the second recognition result.