-
公开(公告)号：EP4425482A2
公开(公告)日：2024-09-04
申请号：EP22909972.6
申请日：2022-12-20
发明人： HUANG, Jiahong , LI, Yule , XIANG, Wei
IPC分类号： G10L13/02
摘要： The present application provides a model training and tone conversion method and apparatus, a device and a medium. By means of a tone extraction network, a first tone feature of input sample audio data can be obtained, so as to obtain tone information of the input sample audio data, which facilitates subsequently obtaining synthesized audio data according to the tone feature, thereby improving the accuracy of the tone of the synthesized audio data. By means of a tone-removing network, and on the basis of the first tone feature, a first semantic feature of the sample audio data can be obtained, thereby accurately obtaining a feature of the sample audio data that is not-related to the tone of the speaker but is related to the spoken content, which facilitates subsequently obtaining synthesized audio data according to the first semantic feature, and ensures the accuracy of the spoken content of the synthesized audio data. After obtaining the trained tone conversion model, tone conversion is carried out by means of the tone conversion model, so that the conversion effect and reliability of tone conversion can be improved.
-
公开(公告)号：EP4290448A1
公开(公告)日：2023-12-13
申请号：EP22749076.0
申请日：2022-01-28
发明人： LI, An , LI, Yule , XIANG, Wei
IPC分类号： G06T3/00 , G06T11/00 , G06N3/08 , G06N3/0455 , G06N3/0475 , G06N3/094
摘要： An image generation model training method, a generation method, an apparatus, and a device. The training method comprises: training and obtaining a first transformation model, the first transformation model being used for generating a first training image on the basis of a first noise sample, and the first training image being an image of a first style (101); training and obtaining a reconstruction model on the basis of the first transformation model (102); training and obtaining a second transformation model, the second transformation model being used for generating a second training image on the basis of a second noise sample, and the second training image being an image of a second style (103); grafting the first transformation model and the second transformation model and generating a grafted transformation model (104); and generating an image generation model on the basis of the reconstruction model and the grafted transformation model, the image generation model being used for transforming an image to be transformed of the first style into a target image of the second style (105).
-
3.发明公开
公开(公告)号：EP4195149A1
公开(公告)日：2023-06-14
申请号：EP21852495.7
申请日：2021-08-06
发明人： XIANG, Wei , WANG, Jundong , CHEN, Lvran , JIAO, Jiening
IPC分类号： G06T7/20
摘要： A target detection and tacking method and apparatus, an electronic device, and a storage medium. The target detection and tracking method comprises: inputting each video frame received from a video stream into an tracking network to obtain a target tracking result of each video frame (S101); when each video frame is a key video frame, inputting the key frame into a detection network and controlling the detection network to operate during a period of receiving the specified number of delay frames, and obtaining a target detection result output by the detection network after the last delay frame in the specified number of delay frames is received (S102); and generating a final target tracking result according to the target detection result and the target tracking result of the last delay frame (S103).
-
公开(公告)号：EP4068150A1
公开(公告)日：2022-10-05
申请号：EP20893388.7
申请日：2020-08-07
发明人： XIANG, Wei , WANG, Yifeng
摘要： A hand key point detection method, a gesture recognition method, and related devices. The hand key point detection method comprises: obtaining a hand image under detection (S101); inputting the hand image into a pre-trained thermodynamic diagram model to obtain a thermodynamic diagram of hand key points, the thermodynamic diagram comprising two-dimensional coordinates of the hand key points (SI02); inputting the thermodynamic diagram and the hand image into a pre-trained three-dimensional information prediction model to obtain hand structured connection information (S103); and determining three-dimensional coordinates of the hand key points in the world coordinate system according to the hand structured connection information and the two-dimensional coordinates in the thermodynamic diagram (S104).
-
5.发明公开
公开(公告)号：EP4414940A1
公开(公告)日：2024-08-14
申请号：EP22894809.7
申请日：2022-11-16
发明人： LI, An , LI, Yule , XIANG, Wei
IPC分类号： G06T17/00
摘要： A caricaturization model construction method and apparatus, and a device, a storage medium and a program product. The method comprises: generating a preset number of sample true pictures by using a pre-trained first generation model; constructing a second generation model on the basis of the first generation model, and generating, by using the second generative model, a sample cartoon picture corresponding to each sample true picture; combining each sample true picture with the corresponding sample cartoon picture to form a sample image pair, and taking a weight corresponding to the second generation model as an initial weight, fitting a preset initial caricaturization model on the basis of a sample set, which is formed by a plurality of the sample image pairs, so as to generate a caricaturization model for converting a target image into a whole-picture caricaturized image.
-
公开(公告)号：EP4414890A1
公开(公告)日：2024-08-14
申请号：EP22877908.8
申请日：2022-09-30
发明人： LUO, Xiongwen , LU, Jianghu , XIANG, Wei
摘要： The present application discloses a model training and scene recognition method and apparatus, a device, and a medium. When a scene recognition model is trained, training is first performed by means of a first scene label and a standard cross-entropy loss of a sample image to obtain parameters of a core feature extraction layer and a global information feature extraction layer; then, according to a feature map outputted by a local supervised learning (LCS) module having an attention mechanism at each level and a loss value obtained by pixel-by-pixel calculation of the first scene label of the sample image, a weight parameter of the LCS module at each level is trained; and finally, a parameter of a fully-connected decision-making layer of the scene recognition model is obtained by training.
-
公开(公告)号：EP4365841A1
公开(公告)日：2024-05-08
申请号：EP21953048.2
申请日：2021-08-09
发明人： JING, Xue , CHEN, Dejian , CHEN, Jianqiang , CAI, Jiaran , XIANG, Wei
摘要： Embodiments of the present application provide an object pose detection method and apparatus, a computer device, and a storage medium. The method comprises: obtaining image data, the image data comprising a target object; inputting the image data into a two-dimensional detection model, and detecting two-dimensional first pose information when a three-dimensional bounding box is projected to the image data, the bounding box being used for detecting the target object; mapping the first pose information into three-dimensional second pose information; and detecting third pose information of the target object according to the second pose information.
-
公开(公告)号：EP4274236A1
公开(公告)日：2023-11-08
申请号：EP21914277.5
申请日：2021-12-27
发明人： LI, Yiyong , SUN, Zhun , HUANG, Qiushi , JING, Xue , XIANG, Wei
IPC分类号： H04N21/2187 , H04N21/234 , G06K9/00 , G06K9/62 , G06N3/02
摘要： The present application provides a live streaming auditing method and apparatus, a server, and a storage medium. The live streaming auditing method comprises: performing, by means of a high-accuracy auditing model and a high-recall auditing model which are cascaded, preliminary violation auditing on the current live streaming frame in a to-be-audited live streaming room; and if the current live streaming frame passes the preliminary violation auditing, inputting, into a pre-constructed behavior auditing model, a first violation score of the current live streaming frame under the high-accuracy auditing model, a second violation score of the current live streaming frame under the high-recall auditing model, and a multi-dimensional behavior feature of the current live streaming frame in said live streaming room, to obtain a target violation score of the current live streaming frame.
-
公开(公告)号：EP4198775A1
公开(公告)日：2023-06-21
申请号：EP21874154.4
申请日：2021-08-30
发明人： LI, Yiyong , HUANG, Qiushi , SUN, Zhun , JING, Xue , XIANG, Wei
IPC分类号： G06F17/18 , G06F16/9536 , G06F16/2458
摘要： An abnormal user auditing method and apparatus, an electronic device, and a storage medium. The abnormal user auditing method comprises: acquiring historical behavior data of a plurality of users to be audited, wherein the historical behavior data comprises historical behavior data formed by interacting between said plurality of users (S101); extracting a plurality of predetermined effective features from the historical behavior data for each said user, wherein the predetermined effective features are features in predetermined sample data (S102); calculating, according to a predetermined probability of an event associated with the plurality of effective features, a probability that each said user is an abnormal user (S103); establishing a total probability function by using a probability that said plurality of users are abnormal users (S104); solving a maximum value of the total probability function by using a predetermined condition as a constraint, to determine candidate users (S105); and auditing the candidate users to obtain an abnormal user (S106).
-
10.发明公开
公开(公告)号：EP4068160A1
公开(公告)日：2022-10-05
申请号：EP20892477.9
申请日：2020-08-20
发明人： XIANG, Wei , PEI, Chao
IPC分类号： G06N3/04
摘要： Disclosed are a neural network training and method for detecting a face and apparatus, and a device and a storage medium. The training method comprises: determining a neural network; training the neural network at a first learning rate according to a first optimization mode, the first learning rate being updated when the neural network is trained each time; mapping the first learning rate of the first optimization mode into a second learning rate of a second optimization mode in the same vector space; determining that the second learning rate satisfies a preset update condition; and continuing to train the neural network at the second learning rate according to a second optimization mode.
-
-
-
-
-
-
-
-
-
搜索结果
10 条记录 (耗时 0.013 秒)
