Author(s)
Gianna McLeod, Sarah Tanveer, and Ali Abedi
Abstract
Accurate indoor–outdoor detection plays a crucial role in dynamic spectrum sharing and interference management, yet existing methods often lack reproducibility and scalability. Inspired by David Donoho’s principle of frictionless reproducibility—where progress accelerates when data, code, and competition are all openly available—we developed a platform to bring these principles into spectrum research. Our system provides a complete pipeline for indoor–outdoor detection: open datasets collected from signals of opportunity, baseline algorithms, and an automated benchmarking infrastructure. Built on HuggingFace, the platform enables researchers to upload their models, have them evaluated automatically, and view results on a public leaderboard. This creates a transparent and competitive environment for testing approaches, similar to what catalyzed rapid advances in AI/ML. By lowering barriers to entry and enabling reproducibility, our framework provides a foundation for accelerating innovation in spectrum research, while also serving as a valuable teaching and community-building tool.