The official implementation for the paper "Iterative Sound Source Localization for Unknown Number of Sources".
See https://github.com/FYJNEVERFOLLOWS/VCTK-SIM
-
torch+torchaudio <https://pytorch.org/>
-
apkit <https://github.com/hwp/apkit>
_ (version 0.2) -
scikit-learn
-
scipy
- Run train_ssnet.py and save a checkpoint of SSNet;
- Load the saved SSNet's path and run inference_sps.py on another training set to build the spatial spectrum dataset;
- Run train_asdnet.py on the spatial spectrum dataset and save a checkpoint of ASDNet;
- Load both SSNet and ASDNet, run inference.py to test DOA estimation performance on unknown multiple sources.