公开(公告)号：WO2021262603A1

公开(公告)日：2021-12-30

申请号：PCT/US2021/038262

申请日：2021-06-21

申请人： NVIDIA CORPORATION

发明人： PARK, Minwoo ,  KWON, Junghyun ,  KOCAMAZ, Mehmet K. ,  SEO, Hae-Jong ,  RODRIGUEZ HERVAS, Berta ,  CHOE, Tae Eun

IPC分类号： G06K9/00 ,  G06K9/46 ,  G06K9/62 ,  G06K9/68 ,  G06K9/20 ,  B60W2554/4029 ,  B60W2554/4044 ,  B60W2556/35 ,  B60W60/00272 ,  G01S13/862 ,  G01S13/865 ,  G01S13/867 ,  G01S13/931 ,  G01S17/931 ,  G01S2013/93271 ,  G01S2015/938 ,  G06K9/6288 ,  G06K9/6293 ,  G06N3/0445 ,  G06N3/0454 ,  G06N3/0481 ,  G06N3/084 ,  G06T2207/20081 ,  G06T2207/20084 ,  G06T7/292 ,  G06V10/16 ,  G06V10/454 ,  G06V20/56 ,  G06V20/58 ,  G06V20/588 ,  G06V30/2552

摘要： In various examples, a multi-sensor fusion machine learning model – such as a deep neural network (DNN) – may be deployed to fuse data from a plurality of individual machine learning models. As such, the multi-sensor fusion network may use outputs from a plurality of machine learning models as input to generate a fused output that represents data from fields of view or sensory fields of each of the sensors supplying the machine learning models, while accounting for learned associations between boundary or overlap regions of the various fields of view of the source sensors. In this way, the fused output may be less likely to include duplicate, inaccurate, or noisy data with respect to objects or features in the environment, as the fusion network may be trained to account for multiple instances of a same object appearing in different input representations.

公开(公告)号：WO2022251697A1

公开(公告)日：2022-12-01

申请号：PCT/US2022/031435

申请日：2022-05-27

申请人： NVIDIA CORPORATION

发明人： RODRIGUEZ HERVAS, Berta ,  DOU, Hang ,  CHEN, Hsin-I ,  ZOU, Kexuan ,  ASSAF, Nizar Gandy ,  PARK, Minwoo

IPC分类号： G06V10/74 ,  G06V10/82 ,  G06V20/58 ,  G06V20/56

摘要： In various examples, lanes may be grouped and a sign may be assigned to a lane in a group, then propagated to another lane in the group to associate semantic meaning corresponding to the sign with the lanes. The sign may be assigned to the most similar lane as quantified by a matching score subject to the lane meeting any hard constraints. Propagation of an assignment of the sign to a different lane may be based on lane attributes and/or sign attributes. Lane attributes may be evaluated and assignments of signs may occur for a lane as a whole, and/or for particular segments of a lane (e.g., of multiple segments perceived by the system). A sign may be a compound sign that is identified as individual signs, which are associated with one another. Attributes of the compound sign may provide semantic meaning used to operate a machine.

公开(公告)号：WO2020185779A1

公开(公告)日：2020-09-17

申请号：PCT/US2020/021894

申请日：2020-03-10

申请人： NVIDIA CORPORATION

发明人： SAJJADI MOHAMMADABADI, Sayed Mehdi ,  RODRIGUEZ HERVAS, Berta ,  DOU, Hang ,  TRYNDIN, Igor ,  NISTER, David ,  PARK, Minwoo ,  CVIJETIC, Neda ,  KWON, Junghyun ,  PHAM, Trung

IPC分类号： G06K9/00 ,  G06K9/32 ,  G06K9/62

摘要： In various examples, live perception from sensors of a vehicle may be leveraged to detect and classify intersections in an environment of a vehicle in real-time or near real-time. For example, a deep neural network (DNN) may be trained to compute various outputs - such as bounding box coordinates for intersections, intersection coverage maps corresponding to the bounding boxes, intersection attributes, distances to intersections, and/or distance coverage maps associated with the intersections. The outputs may be decoded and/or post-processed to determine final locations of, distances to, and/or attributes of the detected intersections.

公开(公告)号：WO2020264029A1

公开(公告)日：2020-12-30

申请号：PCT/US2020/039430

申请日：2020-06-24

申请人： NVIDIA CORPORATION

发明人： PHAM, Trung ,  RODRIGUEZ HERVAS, Berta ,  PARK, Minwoo ,  NISTER, David ,  CVIJETIC, Neda

IPC分类号： G06K9/00 ,  G06K9/46 ,  B60W60/00 ,  G06K9/34 ,  G06K9/44 ,  G06K9/62

摘要： In various examples, live perception from sensors of a vehicle may be leveraged to detect and classify intersection contention areas in an environment of a vehicle in real-time or near real-time. For example, a deep neural network (DNN) may be trained to compute outputs – such as signed distance functions – that may correspond to locations of boundaries delineating intersection contention areas. The signed distance functions may be decoded and/or post-processed to determine instance segmentation masks representing locations and classifications of intersection areas or regions. The locations of the intersections areas or regions may be generated in image-space and converted to world-space coordinates to aid an autonomous or semi-autonomous vehicle in navigating intersections according to rules of the road, traffic priority considerations, and/or the like.

公开(公告)号：WO2020190880A1

公开(公告)日：2020-09-24

申请号：PCT/US2020/022997

申请日：2020-03-16

申请人： NVIDIA CORPORATION

发明人： LEE, Dongwoo ,  KWON, Junghyun ,  OH, Sangmin ,  ZHENG, Wenchao ,  SEO, Hae-Jong ,  NISTER, David ,  RODRIGUEZ HERVAS, Berta

IPC分类号： G06K9/00 ,  G06K9/46 ,  G06K9/62

摘要： A neural network may be used to determine corner points of a skewed polygon (e.g., as displacement values to anchor box corner points) that accurately delineate a region in an image that defines a parking space. Further, the neural network may output confidence values predicting likelihoods that corner points of an anchor box correspond to an entrance to the parking spot. The confidence values may be used to select a subset of the corner points of the anchor box and/or skewed polygon in order to define the entrance to the parking spot. A minimum aggregate distance between corner points of a skewed polygon predicted using the CNN(s) and ground truth corner points of a parking spot may be used simplify a determination as to whether an anchor box should be used as a positive sample for training.