Title:
METHOD FOR DETERMINING A POSE OF AN OBJECT IN THE SURROUNDINGS OF THE OBJECT BY MEANS OF MULTI-TASK LEARNING, AND CONTROL MEANS
Document Type and Number:
WIPO Patent Application WO/2019/137758
Kind Code:
A1
Abstract:
The invention relates to a method for determining a pose of an object in the surroundings thereof, the object and the surroundings thereof being detected as a current image (xi) by means of an optical detection device and the pose of the object being determined by means of an optical image analysis, and the pose of the object being ascertained by means of a neural network (1) as an output of the neural network (1), the neural network (1) being taught by means of multi-task learning (LMTL) by the use of pose regression (Lpose) and descriptor learning (Ld), which descriptor learning is determined by using a triplet-wise loss function (Ltriplet) and a pair-wise loss function (Lpair), the pose regression (Lpose) being determined by means of quaternions, the triplet-wise loss function (Ltriplet) being determined independently of a dynamic margin term (m), and the pair-wise loss function (Lpair) being determined only as an anchoring function.
Inventors:
BUI MAI (DE)
ZAKHAROV SERGEY (DE)
ALBARQOUNI SHADI (DE)
ILIC SLOBODAN (DE)
ZAKHAROV SERGEY (DE)
ALBARQOUNI SHADI (DE)
ILIC SLOBODAN (DE)
Application Number:
PCT/EP2018/085460
Publication Date:
July 18, 2019
Filing Date:
December 18, 2018
Export Citation:
Assignee:
SIEMENS AG (DE)
International Classes:
G06T7/70
Other References:
SERGEY ZAKHAROV ET AL: "3D object instance recognition and pose estimation using triplet loss with dynamic margin", 2017 IEEE/RSJ INTERNATIONAL CONFERENCE ON INTELLIGENT ROBOTS AND SYSTEMS (IROS), 25 September 2017 (2017-09-25), pages 552 - 559, XP055435885, ISBN: 978-1-5386-2682-5, DOI: 10.1109/IROS.2017.8202207
BUI MAI ET AL: "X-Ray PoseNet: 6 DoF Pose Estimation for Mobile X-Ray Devices", 2017 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), IEEE, 24 March 2017 (2017-03-24), pages 1036 - 1044, XP033096888, DOI: 10.1109/WACV.2017.120
WOHLHART PAUL ET AL: "Learning descriptors for object recognition and 3D pose estimation", 2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), IEEE, 7 June 2015 (2015-06-07), pages 3109 - 3118, XP032793759, DOI: 10.1109/CVPR.2015.7298930
BALNTAS VASSILEIOS ET AL: "Pose Guided RGBD Feature Learning for 3D Object Pose Estimation", 2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), IEEE, 22 October 2017 (2017-10-22), pages 3876 - 3884, XP033283259, DOI: 10.1109/ICCV.2017.416
KEHL ET AL., DEEP LERANING OF LOCAL RGB-D PATCHES OF 3D OBJECT DETECTION AND 6D POSE ESTIMATION
BUI MAI ET AL: "X-Ray PoseNet: 6 DoF Pose Estimation for Mobile X-Ray Devices", 2017 IEEE WINTER CONFERENCE ON APPLICATIONS OF COMPUTER VISION (WACV), IEEE, 24 March 2017 (2017-03-24), pages 1036 - 1044, XP033096888, DOI: 10.1109/WACV.2017.120
WOHLHART PAUL ET AL: "Learning descriptors for object recognition and 3D pose estimation", 2015 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), IEEE, 7 June 2015 (2015-06-07), pages 3109 - 3118, XP032793759, DOI: 10.1109/CVPR.2015.7298930
BALNTAS VASSILEIOS ET AL: "Pose Guided RGBD Feature Learning for 3D Object Pose Estimation", 2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION (ICCV), IEEE, 22 October 2017 (2017-10-22), pages 3876 - 3884, XP033283259, DOI: 10.1109/ICCV.2017.416
KEHL ET AL., DEEP LERANING OF LOCAL RGB-D PATCHES OF 3D OBJECT DETECTION AND 6D POSE ESTIMATION
Download PDF:
Previous Patent: TURBINE AIRFOIL, TURBINE BLADE AND METHOD FOR THE PRODUCTION THEREOF
Next Patent: VACUUM INTERRUPTER
Next Patent: VACUUM INTERRUPTER