Cosentino/ConvTasNet_LibriMix_sep_clean

Description:

This model was trained by Joris Cosentino using the librimix recipe in Asteroid. It was trained on the sep_clean task of the LibriMix  dataset.

Training config:

• data:
  • n_src: 2
  • sample_rate: 8000
  • segment: 3
  • task: sep_clean
  • train_dir: data/wav8k/min/train-360
  • valid_dir: data/wav8k/min/dev
• filterbank:
  • kernel_size: 16
  • n_filters: 512
  • stride: 8
• masknet:
  • bn_chan: 128
  • hid_chan: 512
  • mask_act: relu
  • n_blocks: 8
  • n_repeats: 3
  • 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: 2

Results:

• si_sdr: 14.764543634468069
• si_sdr_imp: 14.764029375607246
• sdr: 15.29337970745095
• sdr_imp: 15.114146605113111
• sir: 24.092904661115366
• sir_imp: 23.913669683141528
• sar: 16.06055906916849
• sar_imp: -51.980784441287454
• stoi: 0.9311142440593033
• stoi_imp: 0.21817376142710482

License notice:

This work "ConvTasNet_Librisep_cleanMix_sepclean" is a derivative of LibriSpeech ASR corpus by Vassil Panayotov, used under CC BY 4.0. "ConvTasNet_Librisep_cleanMix_sepclean" is licensed under Attribution-ShareAlike 3.0 Unported by Cosentino Joris.