An open-source robust machine learning platform for real-time detection and classification of 2D material flakes

Jan-Lucas Uslu; Taoufiq Ouaj; David Tebbe; Alexey Nekrasov; Jo Henri Bertram; Marc Schütte; Kenji Watanabe; Takashi Taniguchi; Bernd Beschoten; Lutz Waldecker; Christoph Stampfer

doi:10.1088/2632-2153/ad2287

ジャーナル論文 An open-source robust machine learning platform for real-time detection and classification of 2D material flakes

; ; ; ; ; ;

;

; ; ;

Uslu_2024_Mach._Learn.-_Sci._Technol._5_015027.pdf

Download file (4.81MB)
Download as zip (4.61MB)

コレクション

引用

Jan-Lucas Uslu, Taoufiq Ouaj, David Tebbe, Alexey Nekrasov, Jo Henri Bertram, Marc Schütte, Kenji Watanabe, Takashi Taniguchi, Bernd Beschoten, Lutz Waldecker, Christoph Stampfer. An open-source robust machine learning platform for real-time detection and classification of 2D material flakes. Machine Learning: Science and Technology. 2024, 5 (1), 015027. https://doi.org/10.1088/2632-2153/ad2287

(BibTeX)

説明:

(abstract)

The most widely used method for obtaining high-quality two-dimensional materials is through mechanical exfoliation of bulk crystals. Identifying suitable flakes from a random distribution of crystal thicknesses and sizes on a substrate is typically done manually, which is time-consuming and tedious and suitable flakes can easily be overlooked. Here, we present a platform for fully automated scanning, detection, and classification of two-dimensional materials, the source code of which we make openly available. Our platform is designed to be accurate, reliable and fast as well as versatile in terms of integrating new materials, making it suitable for everyday laboratory work. The implementation allows a fully automized scanning and analysis of wafers with an average inference time of 100 ms for 2.3 MPixel images. The developed detection algorithm is based on a combination of the flakes’ optical contrast towards the substrate and their geometric shape. We demonstrate that it is able to detect the majority of exfoliated flakes of various materials, with an average recall (AR50) of different materials between 66% and 92%. We also show that the algorithm can be trained with as few as five flakes of a given material, which we demonstrate for the examples of few layer graphene, WSe2, CrI3, 1T-TaS2 and hBN. Additionally, we discuss the stability of the approach to variations in the oxide thickness of the wafers used for exfoliation. Our platform has been tested over a period of two years, during which over 106 images of multiple different materials were acquired by over 30 individual researchers.

権利情報: