Research-Driven Development

Over the years, I have contributed to a wide range of academic and industrial projects focused on the analysis of audio signals. These efforts have resulted in the development of numerous benchmark systems and open-source tools addressing challenges in sound classification, event detection, and acoustic scene analysis.

Much of this work has been carried out in the context of the DCASE (Detection and Classification of Acoustic Scenes and Events) community, where I have played a key role in organizing challenges and workshops. My contributions include benchmark systems, evaluation toolboxes, data visualization tools, and a suite of custom plugins for the Pelican static site generator, which powers the DCASE websites.

Beyond DCASE, these tools have proven valuable in broader research and development contexts while supporting reproducibility, benchmarking, and rapid prototyping in machine learning for audio. This page serves as a central catalog of selected tools, systems, and web development projects that have been made publicly available for the community.

Most of the tools are open source and available on GitHub, where you can explore the repositories, contribute to ongoing development, or adapt them for your own research.

Machine Learning

This section features a collection of machine learning systems and tools developed for audio analysis tasks such as acoustic scene classification, sound event detection, and audio tagging. It includes:

Evaluation and Data Handling Tools: Standardized Python libraries such as sed_eval and dcase_util for evaluating system performance and managing audio datasets.
Visualization Tools: Interactive and video-based visualizations using sed_vis for presenting system outputs and annotations, and dynamic data tables and visualizations with js-datatable.
Machine Learning Tutorials: Hands-on code examples from workshops and tutorials (ICASSP 2019 Tutorial about deep learning for acoustic scenes and events, Practical tools for sound classification and speech AI presented in AI Hub Audio and Speech Technology Workshop 2022).
Example Systems from Research: Implementations from the book Computational Analysis of Sound Scenes and Events demonstrating single-label and multi-label classification tasks as well as sound event detection task.
DCASE Challenge Baselines: Reference systems for acoustic scene classification and sound event detection, implemented in Python and MATLAB.

This section is ideal for researchers and developers looking to explore or build upon proven machine learning techniques in the audio domain.

Website Development

This section showcases complete websites developed to support academic and community initiatives:

DCASE Community Website: A central hub for the international DCASE research community, built with Pelican and enhanced with custom plugins for managing citations, personnel, datasets, and events.
bbStat: A grassroots basketball statistics platform for regional leagues in Finland. It features a custom-built Joomla component for managing game data, player stats, and league standings, serving thousands of users annually.

This section demonstrates how web technologies can be applied to both academic and recreational domains with a focus on structured data and user engagement.

Web Development & Utilities

This section contains a suite of Pelican plugins and themes developed to streamline the creation of academic websites. These tools are designed to be modular, reusable, and easy to integrate into static site workflows.

Pelican Plugins for Structured Content: Tools for generating publication lists, personnel directories, file repositories, sponsor listings, and more, all from structured YAML or BibTeX data.
General Content Enhancements: Plugins for generating tables of contents, listing recent articles, and tracking file modification times.
Themes for Academic Sites: Custom Bootstrap-based themes designed for academic communities and personal research websites.

These utilities are actively used in the DCASE website and other academic platforms, and are freely available for integration into your own Pelican-based projects.

GitHub repository

Most of the created toolboxes, libraries, and plugins are open source and maintained on GitHub. You can explore the repositories, contribute to ongoing development, or adapt the tools for your own research and applications.

Github