TASNET: Time-domain Audio Separation Network

Two models are provided here, one with SDR = 11.7 and the other with SDR = 13.6

Model with SDR = 11.7

You can find the oringnal paper here. The testing results are shown as follows:

Function

• 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

Training stage：

• 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

Model with SDR = 13.6

You can find the oringnal paper here.

Function

• 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.