公开(公告)号：US20230177646A1

公开(公告)日：2023-06-08

申请号：US18161123

申请日：2023-01-30

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Songjiang LI ,  Takashi ISOBE ,  Xu JIA ,  QI TIAN

IPC: G06T3/40 ,  G06V10/80 ,  G06V10/77 ,  G06V10/74

CPC classification number: G06T3/4053 ,  G06V10/80 ,  G06V10/761 ,  G06V10/7715

Abstract: An image processing method and apparatus in the field of artificial intelligence, including: decomposing a first image to obtain a first structure sub-image and a first detail sub-image, where the first image is any frame of image in video data other than a first frame; fusing first hidden state information and the first structure sub-image to obtain a second structure sub-image, and splicing the first hidden state information and the first detail sub-image to obtain a second detail sub-image; performing feature extraction based on the second structure sub-image and the second detail sub-image to obtain a structure feature and a detail feature; and obtaining an output image based on the structure feature and the detail feature, where resolution of the output image is higher than resolution of the first image.

公开(公告)号：US20230401830A1

公开(公告)日：2023-12-14

申请号：US18237550

申请日：2023-08-24

Applicant: HUAWEI TECHNOLOGIES CO., LTD.

Inventor： Jianzhong HE ,  Shuaijun CHEN ,  Xu JIA ,  Jianzhuang LIU

IPC: G06V10/774 ,  G06V10/56 ,  G06V10/26 ,  G06N3/045

CPC classification number: G06V10/7753 ,  G06V10/56 ,  G06V10/26 ,  G06N3/045

Abstract: This application provides a model training method in the artificial intelligence field. In a process of determining a loss used to update a model parameter, factors are comprehensively considered. Therefore, an obtained neural network has a strong generalization capability. The method in this application includes: obtaining a first source domain image associated with a target domain image and a second source domain image associated with the target domain image; obtaining a first prediction label of the first source domain image and a second prediction label of the second source domain image through a first to-be-trained model; obtaining a first loss based on the first prediction label and the second prediction label, where the first loss indicates a difference between the first prediction label and the second prediction label; and updating a parameter of the first to-be-trained model based on the first loss, to obtain a first neural network.