Acoupipe - A framework for generating large-scale microphone array data for machine learning AcouPipe is an easy-to-use Python toolbox for generating unique acoustical source localization and characterization data sets with Acoular that can be used for training of deep neural networks and machine learning. Instead of raw time-data, only the necessary input features for acoustical beamforming are stored, which include: Cross-Spectral Matrix / non-redundant Cross-Spectral Matrix Conventional Beamforming Map This allows the user to create datasets of manageable size that are portable and facilitate reproducible research. AcouPipe supports distributed computation with Ray and comes with a default configuration data set inside a pre-built Docker container that can be downloaded from DockerHub. What's Changed in v24.04 New features: Datasets (DatasetSynthetic and DatasetMIRACLE): include a new feature targetmap_analytic and targetmap_estimated, which is a sparse mapping of the analytic / estimated squared sound pressure distribution. "Estimated" means from a limited number of snapshots (e.g. via Welch's method) (feature by @adku1173 in https://github.com/adku1173/acoupipe/pull/34) Bugfixes Datasets: sample the squared RMS value as at source strength instead of the pure RMS @adku1173 in https://github.com/adku1173/acoupipe/pull/35 Others uses an updated version of the MIRACLE dataset AcouPipe now requires at least Acoular 24.03 Full Changelog: https://github.com/adku1173/acoupipe/compare/v23.11...v24.04