Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe
O: Fachverband Oberflächenphysik
O 28: Organic Molecules at Surfaces 3: Theory
O 28.6: Vortrag
Dienstag, 6. September 2022, 11:45–12:00, S051
Classifying Chiral Structure by a Convolutional Neural Network — •Peer Kasten1, Mandy Stritzke2, Johannes Tim Seifert1, Björn Möller3, Timo de Wolff2, Tim Fingscheidt3, and Uta Schlickum1 — 1Institut für Angewandte Physik, Technische Universität Braunschweig — 2Institut für Analysis und Algebra, Technische Universität Braunschweig — 3Institut für Nachrichtentechnik, Technische Universität Braunschweig
Scanning tunneling microscopy (STM) is an important tool to image surfaces at atomic scale. To examine structures of molecules in STM images can be a difficult and time-consuming task. We present a method to recognize chirality within experimentally observed self-assembled molecular structures using the convolutional neural network (CNN) based object detection framework Faster R-CNN. Thereby we can classify unit cells in the image towards one of both chiral structures.
To train the neural network, a sufficient amount of correctly labeled images is necessary. To obtain such data and labels, we utilize a method to create realistic-looking, synthetic STM images in varying zoom-sizes containing lifelike properties such as noise and step edges along with corresponding labels.
Using this synthetic data, we trained a model capable of classifying synthetic images at sizes ranging from 8 nm to 100 nm with high performance. Evaluations of the CNN’s predictions for real images show that this network trained on synthetic data can generalize towards inference on real images.