-
公开(公告)号：EP4099213A1
公开(公告)日：2022-12-07
申请号：EP21176982.3
申请日：2021-05-31
IPC分类号： G06K9/00
摘要： A method for training a convolutional neural network to deliver an identifier of a person visible on an image when the image is inputted to the convolutional neural network, the method comprising:
a. obtaining a training dataset including images of a person and skeleton representations,
b. obtaining feature maps from the convolutional neural network,
c. extracting features using the skeleton representations,
d. forming graphs using the features and processing them in a graph convolutional neural network,
e. calculating a loss,
f. jointly training the two neural networks.-
公开(公告)号：EP4099226A1
公开(公告)日：2022-12-07
申请号：EP21176983.1
申请日：2021-05-31
摘要： A training method for training a CVAE to calculate a prediction Y relative to future properties of agent(s) Ai, and a prediction method to calculate such predictions using the CVAE.
The CVAE comprises a posterior (Po), a prior (Pr) and a decoder (D).
Each of the posterior (Po) and the prior (Pr) comprises an encoder (PoE), a sampler (PoS) and an attention mechanism (PoAM,PrAM).
The encoders (PoE,PrE) calculate parameters of conditional distributions of intermediate variables (PoW,PrW), based on past trajectories of the agents.
The attention mechanisms (PoAM,PrAM) output values of the latent space variable Z based on the drawn value of the intermediate variable (PoW,PrW).
The decoder (D) calculates predictions Y based on a value of the latent space variable Z.
Both the posterior distribution q φ (Z|X,Y) and the prior distribution p θ (Z|X) are joint distributions based on the past observations of the agents.
Computer program(s), readable medium, prediction system and method.-
公开(公告)号：EP4120132A1
公开(公告)日：2023-01-18
申请号：EP21185912.9
申请日：2021-07-15
发明人： OLMEDA REINO, Daniel , REZAEIANARAN, Farzaneh , SHETTY, Rakshith , ZHANG, Shanshan , SCHIELE, Bernt
IPC分类号： G06K9/62
摘要： A computer-implemented method for training a detection model according to unsupervised domain adaptation approach, said method comprising a set of steps performed for each image of at least one pair of images, the images of a pair respectively belonging to a source domain and a target domain. Said set of steps associated with an image comprises:
- obtaining (E10) one or more object proposals and feature vectors for said image,
- clustering (E20) the obtained object proposals by executing a clustering algorithm,
- determining (E30), for each obtained cluster, a quantity representative of the feature vectors respectively associated with the object proposals belonging to said cluster.
The method also comprises a step of learning (E40) a domain discriminator using adversarial training, so as to align between the source and target domains the quantities determined for each pair.-
公开(公告)号：EP4455994A1
公开(公告)日：2024-10-30
申请号：EP23170809.0
申请日：2023-04-28
发明人： LI, Zhi , OLMEDA REINO, Daniel , SHI, Shaoshuai , SCHIELE, Bernt , DAI, Dengxin
摘要： A computer-implemented method for test-time domain adaptation of a first depth estimation model pre-trained on source data, the method comprising, for each target image of a set of target data, the steps of: (S10) aligning a scale of the target image to a scale of the source data, thereby generating an aligned image; and (S20) using the first depth estimation model to generate a depth map prediction for the aligned image.
-
公开(公告)号：EP4455940A1
公开(公告)日：2024-10-30
申请号：EP23170508.8
申请日：2023-04-27
IPC分类号： G06N3/0455 , G06N3/088 , G06N3/0895 , G06N3/09
摘要： A computer-implemented method for training a neural network system, comprising: obtaining a first out-of-distribution, OOD, embedding of an OOD input sample that is out of each class in a training distribution comprising one or more classes; determining a value of a first loss function based on similarity or a distance between the first OOD embedding and one of both a prototype of a class in the training distribution and an in-distribution ("ID") embedding of an ID input sample belonging to the one or more classes, wherein the value of the first loss function positively depends on the similarity or negatively depends on the distance; and modifying a parameter of the system to reduce the value of the first loss function
-
公开(公告)号：EP3975062A1
公开(公告)日：2022-03-30
申请号：EP20198215.4
申请日：2020-09-24
IPC分类号： G06N3/08
摘要： A method for selecting data to train a model, said model being parameterized by weights and represented by a classification function F mapping at least one datum to at least one class. Said method comprises the steps of:
- training (F10) the model using a first dataset (E1), so as to obtain a first set (Θ1) of weight values optimising a function L evaluating the performance of the function F,
and, for each datum (x_i) of a second dataset (E2) comprising unlabelled data,
- determining (F20) at least one pseudo-label (y_i_j) of said datum using said function F,
- determining (F30), using said function L and said first set of weight values, a value (S_i) referred to as "criterion value" evaluating, for a third dataset (E3), a classification error of the labels estimated on the third dataset when said at least one pseudo-label determined for said datum of the second dataset is assumed to be true.
Said method further comprises a step of selecting (F40), from said second dataset, a given number K of data whose respectively associated criterion values satisfy a given selection condition.-
公开(公告)号：EP4293583A1
公开(公告)日：2023-12-20
申请号：EP22178858.1
申请日：2022-06-14
摘要： The invention concerns a computer-implemented method for training a classification model, said method comprising the steps of:
- obtaining (S10) a classification model comprising a representation backbone (320) configured to generate a representation of input samples and to group the input samples into clusters according to a similarity criteria of the representations associated to the input samples, the classification model further comprising a linear classifier (330) configured for assigning a vector P1 to a cluster, each component P1[k] of the vector P1 corresponding to an estimate of the probability of the cluster belonging to a class c[k], k ranging from 1 to K;
- jointly training (S20) the representation backbone and the linear classifier by minimizing a loss function which depends on parameters of the representation backbone and weights of the linear classifier; and,
- updating (S30) parameters of the representation backbone and weights of the linear classifier, so as to obtain an updated classification model.-
公开(公告)号：EP4428719A1
公开(公告)日：2024-09-11
申请号：EP23161346.4
申请日：2023-03-10
发明人： CHUMERIN, Nikolay , AL JUNDI, Rahaf , OLMEDA REINO, Daniel , VOJÍR, Tomas , MATAS, Jiri , SOCHMAN, Jan
IPC分类号： G06F18/243 , G06V10/44 , G06N3/02 , G06F18/213 , G06F18/2415
CPC分类号： G06F18/243 , G06V10/454 , G06F18/213 , G06F18/2415 , G06N3/0499 , G06N3/0895 , G06N3/09
摘要： A system for detecting whether an element belongs to at least one class, the system being configured to:
- obtain an in-distribution probability distribution (26ID) of the at least one class, the in-distribution probability distribution being a distribution of a probability to belong to the at least one class as a function of a result of at least one classifier model;
- obtain an out-of-distribution probability distribution (26OOD), the out-of-distribution probability distribution being a distribution of a probability not to belong to any one of the at least one class as a function of the result of the at least one classifier model;
- determine whether the element belongs to the at least one class based on a comparison between a likelihood that the element belongs to one of the at least one class, determined based on the in-distribution probability distribution, and a likelihood that the element does not belong to any one of the at least one class, determined based on the out-of-distribution probability distribution.-
9.发明公开
公开(公告)号：EP4372693A1
公开(公告)日：2024-05-22
申请号：EP22208431.1
申请日：2022-11-18
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , The Chancellor, Masters and Scholars of the University of Cambridge
IPC分类号： G06T17/00
CPC分类号： G06T17/00
摘要： A computer-implemented method for 3D reconstruction is provided. The method includes: receiving an input image depicting an object; encoding, by an encoder including one or more neural networks, the input image to a latent code; transforming, by a transcoder including one or more neural networks, the latent code to one or more 3D codes each controlling an attribute of a mathematical 3D representation of the object.
-
10.发明公开
公开(公告)号：EP4152274A1
公开(公告)日：2023-03-22
申请号：EP21197966.1
申请日：2021-09-21
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , The Chancellor, Masters and Scholars of the University of Cambridge
摘要： A system (10) for predicting an occupancy probability of at least one point (x) in an environment, based on at least one image (12) of said environment, the system (10) comprising an image encoder (20) configured to extract features (22) from the at least one image (12), a detection module (30) configured to detect at least one object in the image (12) and to associate shape information (32) to the detected at least one object, and an implicit function network (50) configured to predict the occupancy probability ( p ( ô )) based on the features (22) and the shape information (32). A related method and a method for training such a system are further disclosed herein.
-
-
-
-
-
-
-
-
-
搜索结果
17 条记录 (耗时 0.014 秒)
