-
CanonPose: Self-Supervised Monocular 3D Human Pose Estimation in the Wild
Authors:
Bastian Wandt,
Marco Rudolph,
Petrissa Zell,
Helge Rhodin,
Bodo Rosenhahn
Abstract:
Human pose estimation from single images is a challenging problem in computer vision that requires large amounts of labeled training data to be solved accurately. Unfortunately, for many human activities (\eg outdoor sports) such training data does not exist and is hard or even impossible to acquire with traditional motion capture systems. We propose a self-supervised approach that learns a single…
▽ More
Human pose estimation from single images is a challenging problem in computer vision that requires large amounts of labeled training data to be solved accurately. Unfortunately, for many human activities (\eg outdoor sports) such training data does not exist and is hard or even impossible to acquire with traditional motion capture systems. We propose a self-supervised approach that learns a single image 3D pose estimator from unlabeled multi-view data. To this end, we exploit multi-view consistency constraints to disentangle the observed 2D pose into the underlying 3D pose and camera rotation. In contrast to most existing methods, we do not require calibrated cameras and can therefore learn from moving cameras. Nevertheless, in the case of a static camera setup, we present an optional extension to include constant relative camera rotations over multiple views into our framework. Key to the success are new, unbiased reconstruction objectives that mix information across views and training samples. The proposed approach is evaluated on two benchmark datasets (Human3.6M and MPII-INF-3DHP) and on the in-the-wild SkiPose dataset.
△ Less
Submitted 30 November, 2020;
originally announced November 2020.
-
Weakly-supervised Learning of Human Dynamics
Authors:
Petrissa Zell,
Bodo Rosenhahn,
Bastian Wandt
Abstract:
This paper proposes a weakly-supervised learning framework for dynamics estimation from human motion. Although there are many solutions to capture pure human motion readily available, their data is not sufficient to analyze quality and efficiency of movements. Instead, the forces and moments driving human motion (the dynamics) need to be considered. Since recording dynamics is a laborious task tha…
▽ More
This paper proposes a weakly-supervised learning framework for dynamics estimation from human motion. Although there are many solutions to capture pure human motion readily available, their data is not sufficient to analyze quality and efficiency of movements. Instead, the forces and moments driving human motion (the dynamics) need to be considered. Since recording dynamics is a laborious task that requires expensive sensors and complex, time-consuming optimization, dynamics data sets are small compared to human motion data sets and are rarely made public. The proposed approach takes advantage of easily obtainable motion data which enables weakly-supervised learning on small dynamics sets and weakly-supervised domain transfer. Our method includes novel neural network (NN) layers for forward and inverse dynamics during end-to-end training. On this basis, a cyclic loss between pure motion data can be minimized, i.e. no ground truth forces and moments are required during training. The proposed method achieves state-of-the-art results in terms of ground reaction force, ground reaction moment and joint torque regression and is able to maintain good performance on substantially reduced sets.
△ Less
Submitted 23 April, 2021; v1 submitted 17 July, 2020;
originally announced July 2020.
-
Demonstration of broadband contrast at 1.2 $λ$/D and greater for the EXCEDE Starlight Suppression System
Authors:
Dan Sirbu,
Sandrine J. Thomas,
Ruslan Belikov,
Julien Lozi,
Eduardo Bendek,
Eugene Pluzhnik,
Dana H. Lynch,
Troy Hix,
Peter Zell,
Olivier Guyon,
Glenn Schneider
Abstract:
The EXoplanetary Circumstellar Environments and Disk Explorer (EXCEDE) science mission concept uses a visible-wavelength Phase-Induced Amplitude Apodization (PIAA) coronagraph to enable high-contrast imaging of circumstellar debris systems and some giant planets at angular separations reaching into the habitable zones of some of the nearest stars. We report on the experimental results obtained in…
▽ More
The EXoplanetary Circumstellar Environments and Disk Explorer (EXCEDE) science mission concept uses a visible-wavelength Phase-Induced Amplitude Apodization (PIAA) coronagraph to enable high-contrast imaging of circumstellar debris systems and some giant planets at angular separations reaching into the habitable zones of some of the nearest stars. We report on the experimental results obtained in the vacuum chamber at the Lockheed Martin Advanced Technology Center in 10% broadband light centered about 650 nm, with a median contrast of $1 \times 10^{-5}$ between 1.2 and 2.0 $λ$/D simultaneously with $3 \times 10^{-7}$ contrast between 2 and 11 $λ/D$ for a single-sided dark hole using a deformable mirror (DM) upstream of the PIAA coronagraph. These results are stable and repeatable as demonstrated by three measurements runs with DM settings set from scratch and maintained on the best 90% out of the 1000 collected frames. We compare the reduced experimental data with simulation results from modeling observed experimental limits; observed performance is consistent with uncorrected low-order modes not estimated by the Low Order Wavefront Sensor (LOWFS). Modeled sensitivity to bandwidth and residual tip/tilt modes is well-matched to the experiment.
△ Less
Submitted 7 September, 2017;
originally announced September 2017.
-
Unconditional entanglement interface for quantum networks
Authors:
Christoph Baune,
Jan Gniesmer,
Sacha Kocsis,
Christina E. Vollmer,
Petrissa Zell,
Jaromír Fiurášek,
Roman Schnabel
Abstract:
Entanglement drives nearly all proposed quantum information technologies. By up-converting part of a 1550 nm two-mode squeezed vacuum state to 532 nm, we demonstrate the generation of strong continuous-variable entanglement between widely separated frequencies. Nonclassical correlations were observed in joint quadrature measurements of the 1550 and 532 nm fields, showing a maximum noise suppressio…
▽ More
Entanglement drives nearly all proposed quantum information technologies. By up-converting part of a 1550 nm two-mode squeezed vacuum state to 532 nm, we demonstrate the generation of strong continuous-variable entanglement between widely separated frequencies. Nonclassical correlations were observed in joint quadrature measurements of the 1550 and 532 nm fields, showing a maximum noise suppression 5.5 dB below vacuum. Our versatile technique combines strong nonclassical correlations, large bandwidth, and in principle, the ability to entangle the telecommunication wavelength of 1550 nm with any optical wavelength.
△ Less
Submitted 12 May, 2016; v1 submitted 2 October, 2015;
originally announced October 2015.
-
The SOFIA Observatory at the Start of Routine Science Operations : Mission capabilities and performance
Authors:
Pasquale Temi,
Pamela M. Marcum,
Erick Young,
Joseph D. Adams,
Sybil Adams,
B. -G. Andersson,
Eric E. Becklin,
Adwin Boogert,
Rick Brewster,
Eric Burgh,
Brent R. Cobleigh,
Steven Culp,
Jim De Buizer,
Edward W. Dunham,
Christian Engfer,
Geoffrey Ediss,
Maura Fujieh,
Randy Grashuis,
Michael Gross,
Edward Harmon,
Andrew Helton,
Douglas Hoffman,
Jeff Homan,
Michael Hutwohl,
Holger Jakob
, et al. (43 additional authors not shown)
Abstract:
The Stratospheric Observatory for Infrared Astronomy (SOFIA) has recently concluded a set of engineering flights for Observatory performance evaluation. These in-flight opportunities are viewed as a first comprehensive assessment of the Observatory's performance and are used to guide future development activities, as well as to identify additional Observatory upgrades. Pointing stability was evalu…
▽ More
The Stratospheric Observatory for Infrared Astronomy (SOFIA) has recently concluded a set of engineering flights for Observatory performance evaluation. These in-flight opportunities are viewed as a first comprehensive assessment of the Observatory's performance and are used to guide future development activities, as well as to identify additional Observatory upgrades. Pointing stability was evaluated, including the image motion due to rigid-body and flexible-body telescope modes as well as possible aero-optical image motion. We report on recent improvements in pointing stability by using an active mass damper system installed on the telescope. Measurements and characterization of the shear layer and cavity seeing, as well as image quality evaluation as a function of wavelength have also been performed. Additional tests targeted basic Observatory capabilities and requirements, including pointing accuracy, chopper evaluation and imager sensitivity. This paper reports on the data collected during these flights and presents current SOFIA Observatory performance and characterization.
△ Less
Submitted 28 May, 2014;
originally announced May 2014.
