专利快速检索-快速检索全球专利，免费商用专利数据库-IPRDB

1. 发明公开

EP2722815A1 OBJECT RECOGNITION DEVICE 审中-公开
标题翻译：对象识别装置
公开(公告)号：EP2722815A1
公开(公告)日：2014-04-23
申请号：EP12800268.0
申请日：2012-06-14
申请人： Toyota Jidosha Kabushiki Kaisha , Eidgenossische Technische Hochschule Zurich
发明人： RAZAVI, Nima , GALL, Jürgen , VAN GOOL, Luc , FUNAYAMA, Ryuji
IPC分类号： G06T7/00
CPC分类号： G06K9/6267 , G06K9/00791 , G06K9/00805 , G06K9/4633 , G06K9/6219 , G06K9/6282 , G06K2209/27
摘要： A learning unit 4 generates a function table indicating the relationship between the class number and position information of an object and the probability of appearance of the object for each small area image pattern of a code book, calculates a sharing matrix indicating the commonality of a feature amount between the classes, makes a tree diagram in which the classes with a similar feature amount are clustered, and calculates the weight of each node in the tree diagram for each small area image pattern. The recognition processing unit 7 compares image data captured by a camera 2 with the code book, selects the closest small area image pattern among a plurality of small area image patterns, extracts the class related to the node with the smallest weight among the nodes with a weight equal to or greater than a threshold value for the selected small area image pattern, and votes the position information of the small area image pattern for the class, thereby recognizing the object.
摘要翻译：学习单元4针对代码簿的每个小区域图像图案生成指示对象的类别编号和位置信息与对象的出现概率之间的关系的函数表，计算指示特征的共性的共享矩阵量之间的量，制作树图，其中具有相似特征量的类被聚类，并计算每个小区域图像图案的树状图中每个节点的权重。识别处理单元7将由照相机2捕获的图像数据与码本进行比较，从多个小区域图像图案中选择最近的小区域图像图案，并将与节点中具有最小权重的节点有关的类别与对于所选择的小区域图像图案，选择等于或大于阈值的权重，并且针对该类别投票小区域图像图案的位置信息，由此识别该目标。

2. 发明公开

EP4231206A1 METHOD FOR MULTI OBJECT TRACKING WITH QUANTUM ADIABATIC COMPUTING 审中-实审
公开(公告)号：EP4231206A1
公开(公告)日：2023-08-23
申请号：EP22157100.3
申请日：2022-02-16
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich - The Swiss Federal Institute of Technology Zurich
发明人： ABBELOOS, Wim , ZÄCH, Jan-Nico , LINIGER, Alexander , DANELLJAN, Martin , DAI, Dengxin , VAN GOOL, Luc
IPC分类号： G06N10/60
摘要： A computer-implemented method for grouping detected objects (O 1 -O D ) such that a group (Gr t ) comprises detected objects (O d ) corresponding to a given type of object, a said detected object (O d ) being on a frame (F i ) belonging to a sequence of frames (F 1 -F N ), said method consisting in determining values (Oval) of assignments (As 11 -As DG ) of said detected objects (O 1 -O D ) to said groups (Gr 1 -Gr G ), said method comprising the following steps:
- obtaining (EOC), for each pair of detected objects (O di , O dj ), a similarity score (Sim didj );
- determining (EDQ), based on said similarity scores (Sim didj ), an objective function (QCF) quadratic with respect to said assignments;
- evaluating (EOp) values of said assignments, said evaluating step comprising the following substeps:
(i) obtaining (EOp10), using a quantum annealing solver (QAS), an estimated solution (Sol) of an unconstrained binary quadratic optimization problem (QUBO) corresponding to said objective function (QCF);
(ii) post-processing (EOp20) said estimated solution (Sol) to obtain said values (Oval) of assignments (AS 11 -As DG ).

3. 发明公开

EP4174775A1 PREDICTION METHOD; COMPUTER PROGRAM, READABLE MEDIUM, SYSTEM AND MOBILE MACHINE FOR IMPLEMENTING THE METHOD 无效
公开(公告)号：EP4174775A1
公开(公告)日：2023-05-03
申请号：EP21205121.3
申请日：2021-10-27
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , Katholieke Universiteit Leuven KU Leuven Research & Development
发明人： ABBELOOS, Wim , KUZNIETSOV, Yevhen , PROESMANS, Marc , VAN GOOL, Luc
IPC分类号： G06T7/579
摘要： Prediction method for iteratively predicting a depth map (DMj), scene description(s) (SDj) and/or a camera motion (CMj-1,j), based on a current image (Ij);
wherein in adaptation iteration(s), the method comprises:
S40) acquiring sample(s) including a sample image (Is) acquired from the camera at a sample time point (ts), and at least one additional image (Ik) acquired from the camera at an additional time point (tk) different from the sample time point (ts);
S50) for each acquired sample, performing preparatory steps in view of adjusting parameters of at least shared layer(s)(SL), said shared layer(s) being shared layer(s) common at least to the scene description estimator (SDE) and to the depth map estimator (DME);
S60) adjusting parameters of at least the shared layer(s) (SL), based on value(s) of an adaptation loss function (L adapt ) calculated for said at least one sample.
A computer program, a readable medium, a system and a mobile machine for implementing the method.

4. 发明公开

EP4361952A1 METHOD AND SYSTEM FOR TRACKING AN OBJECT IN A SCENE SURROUNDING A MOBILE MACHINE 审中-实审
公开(公告)号：EP4361952A1
公开(公告)日：2024-05-01
申请号：EP22203823.4
申请日：2022-10-26
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , Katholieke Universiteit Leuven KU Leuven Research & Development
发明人： ABBELOOS, Wim , OTHMEZOURI, Gabriel , VERBIEST, Frank , DAWAGNE, Bruno , LEMKENS, Wim , PROESMANS, Marc , VAN GOOL, Luc
IPC分类号： G06T7/246 , G06T7/269 , G06T7/277 , G06T7/579
CPC分类号： G06T7/246 , G06T7/269 , G06T7/277 , G06T2207/3024420130101 , G06T2207/3025220130101 , G06T7/579
摘要： A method for tracking a position of an object (100) in a scene (120) surrounding a mobile machine (140) based upon information acquired from monocular images, the method comprising acquiring at least a first image (22) at a first time (t-1) and a second image (24) at a second time (t), the first image and the second image including image data corresponding to the object and a scene feature (125) present in the scene surrounding the mobile machine, detecting the object in the first image (22) and the second image (24), matching features of the scene feature across the first image and the second image, estimating an egomotion of the mobile machine (140) based upon the scene feature matched across the first image and the second image, and predicting a position of the object taking into account the estimation of the egomotion of the mobile machine.

5. 发明公开

EP4332903A1 METHODS FOR COMPRESSING AND DECOMPRESSING DATA USING TENSOR TRAIN DECOMPOSITION, AND ASSOCIATED DEVICES, COMPUTER PROGRAMS AND STORAGE MEDIUM 审中-实审
公开(公告)号：EP4332903A1
公开(公告)日：2024-03-06
申请号：EP22193316.1
申请日：2022-08-31
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich - The Swiss Federal Institute of Technology Zurich
发明人： ABBELOOS, Wim , OBUKHOV, Anton , USVYATSOV, Mikhail , SAKARIDIS, Christos , SCHINDLER, Konrad , VAN GOOL, Luc
IPC分类号： G06T9/00 , G06N3/08
摘要： A computer-implemented method for compressing a multi-dimensional array of data comprising a plurality of vector elements having the same number R of scalar elements, the method comprising obtaining (E10) D subsequent tensor cores (C 1 -C D ) forming a Tensor Train Decomposition, said method comprising at least one iteration comprising following steps of:
• obtaining (E20) current tensor cores;
• obtaining (E30) approximations (A 1 -A B ) of B subsequent vector elements (E 1 -E B ) of said array of data by determining matrix product between sections of said tensor cores at indices indexing said B subsequent vector elements; and
• modifying (E40) values of said tensor cores with gradient descent, to minimize an approximation error between said B subsequent vector elements and their obtained approximations,
and a step of obtaining (E50) compression of said array of data from the Tensor Train Decomposition formed by the tensor cores modified after the plurality of iterations.

6. 发明公开

EP4239579A1 SYSTEMS AND METHODS FOR MONOCULAR DEPTH ESTIMATION 审中-实审
公开(公告)号：EP4239579A1
公开(公告)日：2023-09-06
申请号：EP22159784.2
申请日：2022-03-02
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich - The Swiss Federal Institute of Technology Zurich
发明人： ABBELOOS, Wim , PATIL, Vaishakh , SAKARIDIS, Christos , LINIGER, Alexander , VAN GOOL, Luc
IPC分类号： G06T7/50 , G06N3/04
摘要： Systems and methods estimate depth of 3D points corresponding to pixels in an image using a trained neural network architecture including a pixel-based depth estimator (4) and a seed-pixel-based depth estimator (6). The pixel-based depth estimator (4) predicts plane coefficients of respective pixels in the image, and generates a first depth estimate by converting the plane coefficients to depth values. The seed-pixel-based depth estimator (6) predicts offset vectors between target pixels and seed pixels on planes in the scene represented by the image, resamples the plane coefficients produced by the plane coefficient detector, and generates a second depth estimate by converting the resampled plane coefficients to depth values. The first and second estimates are fused to produce a final depth estimate, notably by taking a weighted average based on a parameter quantifying the confidence that each pixel is on the same plane as its seed pixel.

7. 发明公开

EP4235588A1 NOVEL VIEW GENERATION METHOD, TRAINING METHOD, COMPUTER PROGRAM, READABLE MEDIUM AND SYSTEM 审中-实审
公开(公告)号：EP4235588A1
公开(公告)日：2023-08-30
申请号：EP22158531.8
申请日：2022-02-24
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich - The Swiss Federal Institute of Technology Zurich
发明人： ABBELOOS, Wim , CAI, Shengqu , OBUKHOV, Anton , DAI, Dengxin , VAN GOOL, Luc
IPC分类号： G06T15/08 , G06T15/20
摘要： A computer-implemented novel view generation method for generating a novel view (I gen ) of an object (C) as seen from a novel pose (d), based on an initial view (I real) of the object, comprising:
S20) calculating a latent feature (z) characterizing the object (C), based on the initial view (I real ); and
S40) calculating the novel view (I gen ) of the object using a neural radiance field (NeRF), based on the novel pose (d) and the latent feature (z), the novel view (I gen ) representing the object (C) as seen from the novel pose (d).
Training method for training an encoder (E) to calculate a latent feature (z) based on an initial view (I real) of an object (C), and for training a generator (G) to output a novel view (I gen ) of the object (C) seen from a novel pose (d), based on the latent feature (z) and the novel pose (d).

8. 发明公开

EP4012623A1 EFFICIENT AND STABLE TRAINING OF A NEURAL NETWORK IN COMPRESSED FORM 审中-公开
公开(公告)号：EP4012623A1
公开(公告)日：2022-06-15
申请号：EP20213174.4
申请日：2020-12-10
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich - The Swiss Federal Institute of Technology Zurich
发明人： ABBELOOS, Wim , OBUKHOV, Anton , RAKHUBA, Maksim , HUANG, Zhiwu , GEORGOULIS, Stamatios , DAI, Dengxin , VAN GOOL, Luc
IPC分类号： G06N3/08 , G06N3/063
摘要： A computer-implemented method for training a neural network using a vectorized compute architecture, the neural network having a plurality of layers associated with a respective plurality of weight matrices, comprising: obtaining a data sample from a dataset, the data sample including an input and an expected output of the neural network; and computing a predicted output by applying the neural network, as parameterized by parameterization configurations resulting from processing the plurality of weight matrices, to the data sample, wherein computing the predicted output comprises: reading a batch of parameterization configurations from memory, the read batch comprising parameterization configurations of a plurality of orthogonal matrices; and executing a function expression associated with the read batch to compute simultaneously the plurality of orthogonal matrices.

9. 发明公开

EP3971821A1 IMAGE COMPLETION USING SELF-ATTENTION AND UNCERTAINTY 审中-实审
公开(公告)号：EP3971821A1
公开(公告)日：2022-03-23
申请号：EP20197574.5
申请日：2020-09-22
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , Katholieke Universiteit Leuven KU Leuven Research & Development , ETH Zurich
发明人： ABBELOOS, Wim , OTHMEZOURI, Gabriel , MA, Liqian , GEORGOULIS, Stamatios , VAN GOOL, Luc
IPC分类号： G06T3/40 , H04N5/232
摘要： A system for image completion is disclosed. The system comprises a coordinate generation module configured to receive past frames and a present frame having a first field-of-view (FoV) and to generate a set of coordinate maps, one for each of the received past frames; and a frame aggregation module configured to receive as input the past frames, the present frame, and the coordinate maps and to synthesize, based on said input, a present frame having a second FoV.

10. 发明公开

EP4418215A1 METHOD FOR TRAINING A DOMAIN-ADAPTED NEURAL NETWORK, AND ASSOCIATED DEVICE 审中-实审
公开(公告)号：EP4418215A1
公开(公告)日：2024-08-21
申请号：EP23157035.9
申请日：2023-02-16
申请人： TOYOTA JIDOSHA KABUSHIKI KAISHA , ETH Zurich
发明人： ABBELOOS, Wim , SAKARIDIS, Christos , YU, Fisher , VAN GOOL, Luc
IPC分类号： G06V10/82
CPC分类号： G06V10/82
摘要： The invention concerns a computer-implemented method for training a domain-adapted neural network configured to generate a semantic segmentation of an image, the method comprising: inputting at least one labelled source image belonging to a source domain; at least one unlabeled target image belonging to a target domain different from the source domain; at least one stylized source image; and at least one stylized target image into the neural network; and, training the neural network by minimizing a loss function penalizing differences between features generated from a given image and features generated from the at least one stylized version of the given image.

你已经成功收藏专利！

检索式保存成功!

IPRDB

热门服务

关于我们

友情链接

联系方式