A5081132021- Scientific Computing and Data Management
- Research Data Management Practices
- Mineral Processing and Grinding
- Data Quality and Management
- Brain Tumor Detection and Classification
- Archaeological Research and Protection
- Geological Modeling and Analysis
- 3D Surveying and Cultural Heritage
- Remote-Sensing Image Classification
- Machine Learning in Materials Science
Karlsruhe Institute of Technology
2022-2024
A project has beenstarted to create a distributed testbed based on FDOs and using the DOIP protocol. This paper describes intentions required components build such testbed. It can beseen as first step towards an international FDO
Abstract Thermal Bridges on Building Rooftops (TBBR) is a multi-channel remote sensing dataset. It was recorded during six separate UAV fly-overs of the city center Karlsruhe, Germany, and comprises total 926 high-resolution images with 6927 manually-provided thermal bridge annotations. Each image provides five channels: three color, one thermographic, computationally derived height map channel. The data pre-split into training test subsets suitable for object detection instance segmentation...
Composing training data for Machine Learning applications can be laborious and time-consuming when done manually. The use of FAIR Digital Objects, in which the is machine-interpretable -actionable, makes it possible to automate simplify this task. As an application case, we represented labeled Scanning Electron Microscopy images from different sources as Objects compose a set. In addition some existing services included our implementation (the Typed-PID Maker, Handle Registry, ePIC Data Type...
The application case for implementing and using the FAIR Digital Object (FAIR DO) concept (Schultes Wittenburg 2019), aims to simplify access label information composing Machine Learning (ML) (Awad Khanna 2015) training data. Data sets curated by different domain experts usually have non-identical terms. This prevents images with similar labels from being easily assigned same category. Therefore, them collectively as data in ML comes cost of laborious relabeling. needs be...
Preprocessing data for research, like finding, accessing, unifying or converting, takes up to large parts of research time spans (Wittenburg and Strawn 2018). The FAIR (Findability, Accessibility, Interoperability, Reusability) principles (Wilkinson 2016) aim support facilitate the (re)use data, will contribute alleviating this problem. A Digital Object (FAIR DO) captures resources all kinds (raw metadata, software, ...) in order align them with principles. Objects are expressive,...