Two models are provided here, one with SDR = 11.7 and the other with SDR = 13.6
You can find the oringnal paper here. The testing results are shown as follows:
- dataset.py: read the data into the model
- Tasnet_model.py:the forword network
- Tasnet_train.py:the main function to run
- trainer.py:calculate the loss and for training and testing
- utils.py: process the raw audio and other useful functions
- train.yaml: all the parameters used in the model
- test.py:separate the mixed audio and calculate SDR
- from the beginning: remove the line with "trainer.rerun" in Tasnet_train.py, use "trainer.run" instead
- from a trained model: remove the line with "trainer.run" in Tasnet_train.py,use "trainer.rerun" instead, and change the "model_path" in train.yaml/temp
You can find the oringnal paper here.
- Tasnet_model_13.6.py: the forword network of SDR = 13.6 system
The only diffierence between 11.7 and 13.6 system is the forward network, just replace the Tasnet_model.py with Tasnet_model_13.6.py and run.