EUVIMEDEUVIMED entdecken
Crossref

EMD 1.0 and `emdfile`: an HDF5 / Python Interface

Benjamin H Savitzky, Steven E Zeltmann, Alexandra Bruefach, Alexander Rakowski, Mary Scott, Matthew L Henderson, Colin Ophus

Microscopy and Microanalysis · 22.07.2023 · 29 (Supplement_1) · 721-723

DOI: 10.1093/micmic/ozad067.356 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Microscopy and Microanalysis, online erschienen am 22.07.2023. Verfasst von Benjamin H Savitzky und 6 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 3 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The EMD version 0 (“Berkeley”) file format was defined in 2012 by Colin Ophus as an HDF5 container holding an N-dimensional array that carries its own calibration with it, in the form of dimension vectors. Here, we introduce the next iteration of this format: EMD version 1.0. We additionally introduce `emdfile`, a Python package which interfaces between the EMD 1.0 file spec and Python runtimes – see Figure 1. The updated EMD format is shown diagrammatically in Figure 2, and the full specification is available at [1]. In computer science, a directory tree consists of a root, a set of nodes called files or folders, and a set of paths connecting them. EMD files consist of directory-like trees in which each node represents a block of data. The contained data may be one of several types, each with some self-describing metadata (e.g. dimension vectors for arrays), plus arbitrary additional dictionary-like metadata. Each EMD file may contain any number of trees and each tree may contain any number of nodes with any simple topology. `emdfile` is a small, pip-installable Python package. It reads and writes EMD 1.0 files, and provides a set of classes, instances of which contain data corresponding the EMD 1.0 nodes. The read/write functions can move single nodes, whole trees, or specified tree branches between HDF5 files and `emdfile` class instances in a Python runtime. In a Python session, any object in a tree may be accessed from any other object in the same tree – see Figure 3. The read interface is shown in Figure 4. `emdfile` is designed to be lightweight and modular. It has three user-facing functions and eight classes. It has full unit test coverage, dependencies of h5py, tqdm, and numpy only, and is comprised of 4,500 lines of code (included test). Its full suite of read/write capabilities can be added to existing Jupyter notebooks, scripts, modules, and packages with a minimum of new node. Version py4DSTEM v14+ is built on top of `emdfile` [2]. It is open-source GPLv3 licensed and on github at [3]. Context and use-case diagrams for EMD. EMD interfaces between Python runtime objects and HDF5 storage, and enables revisiting and extending analyses. The EMD 1.0 file specification diagram. EMD 1.0 files contain any number of directory-like trees. Each node contains any number of metadata dictionaries, plus a block of data according to its type. In a python runtime environment, each of the data-holding `emdfile` class instances may be thought of as a tree node. Any node may be accessed from any other node in the same tree. `emdfile` provides read and write functions that move data between runtime trees and EMD 1.0 (HDF5) trees. This diagram shows the read interface.

Zitieren

APA

Savitzky, B. H., Zeltmann, S. E., Bruefach, A., Rakowski, A., Scott, M., Henderson, M. L., & Ophus, C. (2023). EMD 1.0 and `emdfile`: an HDF5 / Python Interface. Microscopy and Microanalysis, 29(Supplement_1), 721-723. https://doi.org/10.1093/micmic/ozad067.356

Vancouver

Savitzky BH, Zeltmann SE, Bruefach A, Rakowski A, Scott M, Henderson ML, et al. EMD 1.0 and `emdfile`: an HDF5 / Python Interface. Microscopy and Microanalysis. 2023;29(Supplement_1):721-723. doi:10.1093/micmic/ozad067.356

BibTeX
@article{Savitzky2023,
  author = {Savitzky, Benjamin H and Zeltmann, Steven E and Bruefach, Alexandra and Rakowski, Alexander and Scott, Mary and Henderson, Matthew L and Ophus, Colin},
  title = {EMD 1.0 and `emdfile`: an HDF5 / Python Interface},
  journal = {Microscopy and Microanalysis},
  year = {2023},
  volume = {29},
  number = {Supplement_1},
  pages = {721--723},
  doi = {10.1093/micmic/ozad067.356}
}
Kostenlos starten →