Audio Event Detection and Localization with Multitask Regression Network

Source code of the submission "Audio Event Detection and Localization with Multitask Regression Network" ranked 6th in the DCASE 2020 SELD challenge.

How to run:

1. Extract feature bash run_feature_extraction.sh
2. Experiments on the development data bash run_dcase_dev.sh
3. Experiments on the evaluation data bash run_dcase_eval.sh

Environment:

• Python3
• librosa
• Tensorflow GPU 1.x (x >= 9) (for network training and evaluation)

Contact:

Huy Phan

Centre for Digital Music (C4DM) Queen Mary University of London Email: [email protected]

License

MIT © Huy Phan