Data generator for creating synthetic audio mixtures suitable for DCASE Challenge 2022 Task 3
The provided code was tested with Python 3.8 and the following libraries: SoundFile 0.10.3, mat73 0.58, numpy 1.20.1, scipy 1.6.2, librosa 0.8.1.
This repository contains several Python file, which in total create a complete data generation framework.
generation_parameters.py is a separate script used for setting the parameters for the data generation process, including things such as audio dataset, number of folds, mixuture length, etc.db_config.py is a class for containing audio filelists and data parameters from different audio datasets used for the mixture generation.metadata_synthesizer.py is a class for generating the mixture target labels, along with the corresponding metadata and statistics. Information from this class can be further used for synthesizing the final audios.audio_synthesizer.py is a class for synthesizing noiseless audio files containing the simulated mixtures.audio_mixer.py is a class for mixing the generated audio mixtures with background noise and/or interference mixtures.make_dataset.py is the main script in which the whole framework is used to perform the full data generation process.utils.py is an additional file containing complementary functions for other scripts.Moreover, two object files are included in case the database configuration via db_config.py takes too much time:
db_config_fsd.obj is a DBConfig class containing information about the database and files for the FSD50K audioset.db_config_nigens.obj is a DBConfig class containing information about the database and files for the NIGENS audioset.Two exemplary scripts are added:
example_script_DCASE2021.py is a script showing a pipeline to generate data similar to the DCASE2021 dataset.example_script_DCASE2022.py is a script showing a pipeline to generate data similar to the current DCASE2022 dataset.The repository is licensed under the TAU License.