search

fgnt / sms_wsj

SMS-WSJ: Spatialized Multi-Speaker Wall Street Journal database for multi-channel source separation and recognition
MIT License
110 stars 25 forks source link

Update default for id_to_file_name #14

Closed jensheit closed 3 years ago

jensheit commented 3 years ago

updates default value for id_to_file_name to make the default to evaluate on multiple speaker

boeddeker commented 3 years ago

Can you make a grep -r "'{}_0.wav'" and change all positions? I don't see any use case for the postfix _0 (At least not to be the default value).

Should we add an assert (if len(target_speaker) > 0) that id_to_file_name.format(_id, spk) yields different file names for different speakers?

Would something like 

if '{}' in id_to_file_name:
    id_to_file_name_fn = lambda _id, spk: id_to_file_name.format(_id, spk)
else:
    id_to_file_name_fn = lambda _id, spk: id_to_file_name.format(id=_id, spk=spk)

be helpful for a user? It would allow a more verbose version and also a swapped order of id and spk.

Can you add a comment in the config function, that the first {} is the example id and the second the speaker id?