elib
DLR-Header
DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Augmented Autoencoders for Object Orientation Estimation trained on synthetic RGB Images

Sundermeyer, Martin (2017) Augmented Autoencoders for Object Orientation Estimation trained on synthetic RGB Images. Master's. DLR-Interner Bericht. DLR-IB-RM-OP-2017-165, 83 S. (Unpublished)

[img] PDF
39MB

Abstract

Fast and accurate object pose estimation algorithms are crucial for robotic tasks. Despite intensive research, most approaches are not generally applicable on arbitrary object characteristics and dynamic environment conditions. Learning-based methods like Convolutional Neural Networks (CNNs) have proven good generalization properties given sufficient training data. However, annotating RGB images with 3D object orientations is difficult and requires expert knowledge. In this work, a real-time approach for joint 2D object detection and 3D orientation estimation is proposed. First, a CNN-based object detector [45] is used to localize objects in an image plane. In the second step, an Autoencoder (AE) predicts the 3D orientation of the object from the resulting scene crop. The main contribution is a new training method for AEs that allows learning 3D object orientations from synthetic views of a 3D model, dispensing with the need to annotate orientations in real sensor data. The AE is trained to revert augmentations applied to the input and thus becomes robust against irrelevant color changes, background clutter and occlusions. It learns to produce low-dimensional representations of synthetic object orientations which can be compared to the representations of real RGB test data in a k-Nearest-Neighbor (kNN) search. Experiments on the pose annotated dataset T-LESS [23] prove the performance of the approach on different sensors. Finally, the training on synthetic data is shown to be almost on par with the training on real data.

Item URL in elib:https://elib.dlr.de/117228/
Document Type:Monograph (DLR-Interner Bericht, Master's)
Title:Augmented Autoencoders for Object Orientation Estimation trained on synthetic RGB Images
Authors:
AuthorsInstitution or Email of AuthorsAuthors ORCID iD
Sundermeyer, Martinmartin.sundermeyer (at) dlr.dehttps://orcid.org/0000-0003-0587-9643
Date:2017
Refereed publication:No
Open Access:Yes
Gold Open Access:No
In SCOPUS:No
In ISI Web of Science:No
Number of Pages:83
Status:Unpublished
Keywords:Object Pose Estimation, Deep Learning, Synthetic Data, Autoencoder, Object Detection, Simulation to Reality Transfer, Augmentation
Institution:Technische Universität München
Department:Electrical Engineering and Information Technology
HGF - Research field:Aeronautics, Space and Transport
HGF - Program:Space
HGF - Program Themes:Space Technology
DLR - Research area:Raumfahrt
DLR - Program:R SY - Technik für Raumfahrtsysteme
DLR - Research theme (Project):R - Vorhaben Multisensorielle Weltmodellierung
Location: Oberpfaffenhofen
Institutes and Institutions:Institute of Robotics and Mechatronics (since 2013) > Perception and Cognition
Deposited By: Sundermeyer, Martin
Deposited On:19 Dec 2017 15:00
Last Modified:31 Jul 2019 20:14

Repository Staff Only: item control page

Browse
Search
Help & Contact
Information
electronic library is running on EPrints 3.3.12
Copyright © 2008-2017 German Aerospace Center (DLR). All rights reserved.