Software Open Access

brijmohan/ConvTasNet_Libri1Mix_enhsingle

Brij Mohan Lal Srivastava

Description:

This model was trained by Brij Mohan Lal Srivastava using the librimix/ConvTasNet recipe in Asteroid. It was trained on the enh_single task of the Libri1Mix dataset.

 

Training config:

  • data:
    • n_src: 1
    • sample_rate: 8000
    • segment: 3
    • task: enh_single
    • train_dir: data/wav8k/min/train-360
    • valid_dir: data/wav8k/min/dev
  • filterbank:
    • kernel_size: 16
    • n_filters: 512
    • stride: 8
  • main_args:
    • exp_dir: exp/train_convtasnet_vpc
    • help: None
  • masknet:
    • bn_chan: 128
    • hid_chan: 512
    • mask_act: relu
    • n_blocks: 8
    • n_repeats: 3
    • n_src: 1
    • skip_chan: 128
  • optim:
    • lr: 0.001
    • optimizer: adam
    • weight_decay: 0.0
  • positional arguments:
    • training:
      • batch_size: 24
      • early_stop: True
      • epochs: 200
      • half_lr: True
      • num_workers: 4

     

    Results:

    • si_sdr: 14.783675142685572
    • si_sdr_imp: 11.464625198953202
    • sdr: 15.497505907983102
    • sdr_imp: 12.07230150154914
    • sir: inf
    • sir_imp: nan
    • sar: 15.497505907983102
    • sar_imp: 12.07230150154914
    • stoi: 0.9270030254700518
    • stoi_imp: 0.1320547197597893

     

    License notice:

    This work "ConvTasNet_Libri1Mix_enhsingle" is a derivative of LibriSpeech ASR corpus by Vassil Panayotov, used under CC BY 4.0; of The WSJ0 Hipster Ambient Mixtures dataset by Whisper.ai, used under CC BY-NC 4.0 (Research only). "ConvTasNet_Libri1Mix_enhsingle" is licensed under Attribution-ShareAlike 3.0 Unported by Brij Mohan Lal Srivastava.
    Files (20.1 MB)
    Name Size
    model.pth
    md5:977c2a79cbb3f0134291a15cf89f3f16     		20.1 MB Download
    493
    121
    views
    downloads
    All versions This version
    Views 493493
    Downloads 121121
    Data volume 2.4 GB2.4 GB
    Unique views 415415
    Unique downloads 115115
    More info on how stats are collected.
    Indexed in
    Publication date:
    August 3, 2020
    DOI:
    10.5281/zenodo.3970768

    Zenodo DOI Badge

    DOI 

    10.5281/zenodo.3970768

    Markdown 

    [![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.3970768.svg)](https://doi.org/10.5281/zenodo.3970768)

    reStructedText 

    .. image:: https://zenodo.org/badge/DOI/10.5281/zenodo.3970768.svg
   :target: https://doi.org/10.5281/zenodo.3970768

    HTML 

    <a href="https://doi.org/10.5281/zenodo.3970768"><img src="https://zenodo.org/badge/DOI/10.5281/zenodo.3970768.svg" alt="DOI"></a>

    Image URL 

    https://zenodo.org/badge/DOI/10.5281/zenodo.3970768.svg

    Target URL 

    https://doi.org/10.5281/zenodo.3970768

    Keyword(s):
    Asteroid audio source separation Libri1Mix enh_single ConvTasNet pretrained model
    Communities:
    License (for files):
    Creative Commons Attribution Share Alike 3.0 Unported

    Versions

    Version 1 10.5281/zenodo.3970768 Aug 3, 2020
    Cite all versions? You can cite all versions by using the DOI 10.5281/zenodo.3970767. This DOI represents all versions, and will always resolve to the latest one. Read more.

    Share

    Cite as

    Export