Asteroid模型mpariante/ConvTasNet_WHAM_sepclean

从 Zenodo 导入

描述:

该模型是由Manuel Pariente使用wham/ConvTasNet配方在 Asteroid 上训练的。它是在WHAM!数据集的sep_clean任务上进行训练的。

训练配置:

data:
    n_src: 2
    mode: min
    nondefault_nsrc: None
    sample_rate: 8000
    segment: 3
    task: sep_clean
    train_dir: data/wav8k/min/tr/
    valid_dir: data/wav8k/min/cv/
filterbank:
    kernel_size: 16
    n_filters: 512
    stride: 8
main_args:
    exp_dir: exp/wham
    gpus: -1
    help: None
masknet:
    bn_chan: 128
    hid_chan: 512
    mask_act: relu
    n_blocks: 8
    n_repeats: 3
    n_src: 2
    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

结果:

si_sdr: 16.21326632846293
si_sdr_imp: 16.21441705664987
sdr: 16.615180021738933
sdr_imp: 16.464137807433435
sir: 26.860503975131923
sir_imp: 26.709461760826414
sar: 17.18312813480803
sar_imp: -131.99332048277296
stoi: 0.9619940905157323
stoi_imp: 0.2239480672473015

许可声明:

此作品"ConvTasNet_WHAM!_sepclean"是基于 CSR-I (WSJ0) Complete 由 LDC 创作的，仅用于 LDC User Agreement for Non-Members (仅限研究)。“ConvTasNet_WHAM!_sepclean”由Manuel Pariente根据 Attribution-ShareAlike 3.0 Unported 许可证授权。