Summary:
The DreamFAIRer project, developed in the context of the HES-SO Open Research Data (ORD) call dedicated to software and source code, addresses the growing need to align research data and software with established principles of Open Research Data practices and, more specifically, the FAIR principles, which require data to be Findable, Accessible, Interoperable, and Reusable. While these principles were initially defined for data, their extension to research software (FAIR4RS) highlights additional challenges related to executability, dependency management, and reproducibility. In this context, the project focuses on improving the FAIR compliance of an existing research codebase developed for data augmentation in fingerprint-based localization systems. The work builds upon previous research on the ProxyFAUG method, which aims to reduce data collection costs in localization tasks through synthetic data augmentation. Although the associated datasets and code had been made publicly available through an open repository (Zenodo), their initial publication did not fully address all aspects of FAIR and FAIR4RS compliance, particularly regarding reproducibility, alignment with community conventions, interoperability, and explicit documentation of the computational environment and workflow. The main objective of DreamFAIRer was therefore to assess the initial alignment of these artifacts with FAIR and FAIR4RS principles and to implement targeted improvements to enhance their reusability and reproducibility. These improvements include: the provision of an updated and better-documented repository [1] regrouping the software contributions of the ProxyFAUG method; the introduction of a time-based train/validation/test split, better aligned with community conventions, replacing the previously used random split; the provision of a Conda environment (.yml) file to support reproducibility; the clarification of versioning for data, code, and dependencies; and the generalization of data handling by removing hardcoded assumptions and documenting the expected input structure. In addition, the project contributes by providing a fully documented and reproducible implementation of the ProxyFAUG method that can serve as a baseline for future method comparison in fingerprint-based localization. A review of the relevant literature indicates that no directly comparable open implementation of a rule-based augmentation baseline is currently available. As such, this work provides a unique contribution by enabling transparent, reproducible, and evidence-based comparison of methods within this category. In addition to these technical contributions, the project delivers a FAIR-compliant software artifact deposited in an open repository with a persistent identifier (DOI), accompanied by documentation supporting its reuse. The present report, which will be deposited in the institutional repository ArODES, describes the implementation of FAIR and FAIR4RS principles in this context, discusses the initial limitations and subsequent improvements, and reflects on the broader implications for reproducible and reusable research software. Overall, DreamFAIRer contributes to the objectives of the Open Research Data initiative by demonstrating how existing research code can be systematically improved to better align with FAIR and FAIR4RS requirements, thereby enhancing transparency, reproducibility, and long-term usability of scientific outputs.
Help
SWITCH edu-ID
Administration
