-
公开(公告)号：EP4102463A1
公开(公告)日：2022-12-14
申请号：EP20917552.0
申请日：2020-09-10
发明人： QIAO, Yu , YANG, Ze , WANG, Yali , CHEN, Xianyu , LIU, Jianzhuang , YUE, Jun
摘要： An image processing method and a related device are disclosed. The method is applied to the field of artificial intelligence. After obtaining a prior bounding box set of a target image, an execution device constructs a bounding box set based on the prior bounding box set, accordingly determines a correlation between each prior bounding box in the prior bounding box set and each bounding box, and finally determines, based on the correlation, a location and/or a category of a target object included in any prior bounding box. A series of transformation operations are performed on the bounding box set, and a feature of each bounding box is integrated into a feature of a prior bounding box, to increase a degree of differentiation between objects included in the prior bounding box. Therefore, background information around an object in the target image is creatively used to help locate or classify the object. This resolves a problem that object categories in the target image are easily confused in a case of a small sample, and significantly improves detection accuracy.
-
公开(公告)号：EP4181015A1
公开(公告)日：2023-05-17
申请号：EP21849301.3
申请日：2021-07-15
发明人： HAO, Lei , YUE, Jun , XU, Songcen
IPC分类号： G06K9/00
摘要： A handheld object recognition method related to the field of artificial intelligence is provided, including: obtaining location information of each of one or more detected objects in a to-be-recognized image, and obtaining a first label of each detected object, where the location information of each detected object is location information of the detected object in the to-be-recognized image, the first label of the detected object indicates a type of the detected object, and the type of the detected object is used to represent a handheld relationship of the detected object; obtaining a handheld object from the one or more detected objects based on the first label of each detected object, and obtaining location information of the handheld object from the location information of the one or more detected objects; and recognizing the handheld object in the to-be-recognized image based on the location information of the handheld object to obtain a recognition result of the handheld object. According to embodiments of this application, when an object or a hand is obviously blocked, the handheld object can be accurately determined. In this way, the handheld object is recognized.
-
公开(公告)号：EP4113370A1
公开(公告)日：2023-01-04
申请号：EP21775238.5
申请日：2021-03-22
发明人： YUE, Jun , QIAN, Li , XU, Songcen , SHAO, Bin
IPC分类号： G06K9/00
摘要： This application provides a method and apparatus for updating an object recognition model in the field of artificial intelligence. In the technical solution provided in this application, a target image and first voice information of a user are obtained. The first voice information indicates a first category of a target object in the target image. A feature library of a first object recognition model is updated based on the target image and the first voice information. The updated first object recognition model includes a feature of the target object and a first label indicating the first category, and the feature of the target object corresponds to the first label. A recognition rate of an object recognition model can be improved more easily according to the technical solution provided in this application.
-
公开(公告)号：EP4149109A1
公开(公告)日：2023-03-15
申请号：EP21818681.5
申请日：2021-05-29
发明人： SHAO, Bin , YUE, Jun , QIAN, Li , XU, Songcen , HUANG, Xueyan , LIU, Yajiao
摘要： This application provides a video generation method and a related device, and may be applied to the field of image processing and video generation in the field of artificial intelligence. The video generation method includes: receiving a video generation instruction, and obtaining text information and image information in response to the video generation instruction, where the text information includes one or more keywords, and the image information includes N images; obtaining, based on the one or more keywords, an image feature that is in each of the N images and that corresponds to the one or more keywords; and inputting the one or more keywords and image features of the N images into a target generator network to generate a target video, where the target video includes M images, and the M images are images that are generated based on the image features of the N images and that correspond to the one or more keywords. During embodiments of this application, a video is automatically generated on the premise of ensuring richness of video content.
-
公开(公告)号：EP4109393A1
公开(公告)日：2022-12-28
申请号：EP21757807.9
申请日：2021-02-19
发明人： LIU, Lin , YUE, Jun , LIU, Jianzhuang , YUAN, Shanxin
摘要： This application provides a moire removal method and an apparatus, and relates to the field of artificial intelligence, and in particular, to the field of computer vision. The method includes: performing wavelet transform on a to-be-processed image to obtain a first wavelet image (S210); performing moire removal processing on the first wavelet image to obtain a second wavelet image (S220); and performing inverse wavelet transform on the second wavelet image to obtain a repaired image (S230) of the to-be-processed image. In the foregoing method, moires on the to-be-processed image are decomposed onto wavelet images with different frequencies by using wavelet transform, and then moire removal processing is separately performed on each wavelet image, to remove moires with different frequencies. This can more effectively remove moires and reduce loss of original details of the image.
-
公开(公告)号：EP4109335A1
公开(公告)日：2022-12-28
申请号：EP20919863.9
申请日：2020-09-25
发明人： WANG, Shuo , YUE, Jun , LIU, Jianzhuang , TIAN, Qi
IPC分类号： G06K9/62
摘要： This application provides a method for training a classifier, including: obtaining a first training sample, where the first training sample includes a corresponding semantic tag; obtaining a plurality of second training samples, where each of the second training samples includes a corresponding semantic tag; determining a target sample from the plurality of second training samples based on semantic similarities between the first training sample and the plurality of second training samples; and training the classifier based on the first training sample, the target sample, and a semantic similarity between the first training sample and the target sample. Training efficiency and performance of the classifier can be improved by training the classifier based on the semantic similarity. In addition, during feature extraction in the foregoing method, a semantic tag is not used for learning. Therefore, a network structure of a feature extractor does not need to be changed, thereby improving training efficiency of a neural network.
-
公开(公告)号：EP3933693A1
公开(公告)日：2022-01-05
申请号：EP20777500.8
申请日：2020-03-26
发明人： YUE, Jun , LIU, Jianzhuang , XU, Songcen , YAN, Youliang , QIAN, Li
IPC分类号： G06K9/62
摘要： This application discloses an object recognition method and apparatus in the field of artificial intelligence. This application relates to the field of artificial intelligence, and specifically, to the field of computer vision. The method includes: obtaining one or more body regions of a to-be-recognized image; determining a saliency score of each of the one or more body regions; and when a saliency score of a body region A is greater than or equal to a categorization threshold, determining a feature vector of an object in the body region A based on a feature of the object in the body region A, and determining a category of the object in the body region A based on the feature vector of the object in the body region A and a category feature vector in a feature library, where the body region A is any one of the one or more body regions.
-
-
-
-
-
-
搜索结果
7 条记录 (耗时 0.021 秒)
