模型:
cankeles/ConvTasNet_WHAMR_enhsingle_16k
任务:
音频到音频许可:
cc-by-sa-4.0描述:
该模型在被称为WHAMR的修改版本上进行了微调!这个版本的发言人来自有声读物录音,并通过Pedalboard、Spotify添加了混响效果。
初始模型来自这里: https://huggingface.co/JorisCos/ConvTasNet_Libri1Mix_enhsingle_16k
该模型由M. Can Keles使用WHAM配方进行了训练在 Asteroid 中。它在WHAM数据集的enh_single任务上进行了训练。
训练配置:
data: mode: min nondefault_nsrc: null sample_rate: 16000 task: enh_single train_dir: wav16k/min/tr/ valid_dir: wav16k/min/cv/ filterbank: kernel_size: 16 n_filters: 512 stride: 8 main_args: exp_dir: exp/tmp help: null 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: 2 early_stop: true epochs: 10 half_lr: true num_workers: 4
结果:
'sar': 13.612368475881558, 'sar_imp': 9.709316571584433, 'sdr': 13.612368475881558, 'sdr_imp': 9.709316571584433, 'si_sdr': 12.978640274976373, 'si_sdr_imp': 9.161273840297232, 'sir': inf, 'sir_imp': nan, 'stoi': 0.9214516928197306, 'stoi_imp': 0.11657488247668318