Conv-TasNet: Surpassing Ideal Time–Frequency Magnitude Masking for Speech Separation

Tensorflow 2.x (with Keras API) Implementation of the Conv-TasNet (Luo et al., 2019)

Training and Predicting

1. Training example

python main.py --checkpoint=checkpoint --dataset_path=dataset_path

• checkpoint: path to save trained weights of model
• dataset_path: path of MUSDB18
• use python main.py --help to see all options

2. Predicting example

python predict.py --checkpoint=checkpoint --video_id=video_id

• checkpoint: path where trained weights of model is saved
• video_id: video id in youtube
• use python predict.py --help to see all options

License

MIT License

References

1. Conv-TasNet: Surpassing Ideal Time–Frequency Magnitude Masking for Speech Separation (IEEE)
2. Tensorflow implementation by paxbun (GitHub)