Synthetic Image Rendering Solves Annotation Problem in Deep Learning Nanoparticle Segmentation
Authors:
Leonid Mill,
David Wolff,
Nele Gerrits,
Patrick Philipp,
Lasse Kling,
Florian Vollnhals,
Andrew Ignatenko,
Christian Jaremenko,
Yixing Huang,
Olivier De Castro,
Jean-Nicolas Audinot,
Inge Nelissen,
Tom Wirtz,
Andreas Maier,
Silke Christiansen
Abstract:
Nanoparticles occur in various environments as a consequence of man-made processes, which raises concerns about their impact on the environment and human health. To allow for proper risk assessment, a precise and statistically relevant analysis of particle characteristics (such as e.g. size, shape and composition) is required that would greatly benefit from automated image analysis procedures. Whi…
▽ More
Nanoparticles occur in various environments as a consequence of man-made processes, which raises concerns about their impact on the environment and human health. To allow for proper risk assessment, a precise and statistically relevant analysis of particle characteristics (such as e.g. size, shape and composition) is required that would greatly benefit from automated image analysis procedures. While deep learning shows impressive results in object detection tasks, its applicability is limited by the amount of representative, experimentally collected and manually annotated training data. Here, we present an elegant, flexible and versatile method to bypass this costly and tedious data acquisition process. We show that using a rendering software allows to generate realistic, synthetic training data to train a state-of-the art deep neural network. Using this approach, we derive a segmentation accuracy that is comparable to man-made annotations for toxicologically relevant metal-oxide nanoparticle ensembles which we chose as examples. Our study paves the way towards the use of deep learning for automated, high-throughput particle detection in a variety of imaging techniques such as microscopies and spectroscopies, for a wide variety of studies and applications, including the detection of plastic micro- and nanoparticles.
△ Less
Submitted 20 November, 2020;
originally announced November 2020.
Electronic friction coefficients from the atom-in-jellium model for $Z=1-92$
Authors:
Nick Gerrits,
J. Iñaki Juaristi,
Jörg Meyer
Abstract:
The break-down of the Born-Oppenheimer approximation is an important topic in chemical dynamics on metal surfaces. In this context, the most frequently used "work-horse" is electronic friction theory, commonly relying on friction coefficients obtained from density functional theory (DFT) calculations from the early 80s based on the atom-in-jellium model. However, results are only available for a l…
▽ More
The break-down of the Born-Oppenheimer approximation is an important topic in chemical dynamics on metal surfaces. In this context, the most frequently used "work-horse" is electronic friction theory, commonly relying on friction coefficients obtained from density functional theory (DFT) calculations from the early 80s based on the atom-in-jellium model. However, results are only available for a limited set of jellium densities and elements ($Z=1-18$). In this work, these calculations are revisited by investigating the corresponding friction coefficients for the entire periodic table ($Z=1-92$). Furthermore, friction coefficients obtained by including the electron density gradient on the Generalized Gradient Approximation (GGA) level are presented. Finally, we show that spin polarization and relativistic effects can have sizeable effects on these friction coefficients for some elements.
△ Less
Submitted 27 October, 2020; v1 submitted 1 October, 2020;
originally announced October 2020.
