Closed eeric closed 5 years ago
Hi! The TIMIT_labels.npy is a simple python dictionary that has sentence_id as key and an integer as value. The integers (that range between 0 and N_speakers-1) are the labels of our speaker-identification system and tell us which is the speaker that is speaking in each utterance. The construction of this dictionary depends on how the dataset is organized and where the speaker-id information is stored. For TIMIT is very easy. If you consider a certain file (e.g., TIMIT/train/dr3/fcmg0/sx72.wav, the folder "fcmg0" is the speaker. You only have to associate "fcmg0" to a unique integer (from 0 to N_speakers-1), which is different from that used for the other speakers.
Mirco
On Fri, 25 Jan 2019 at 03:46, eeric notifications@github.com wrote:
how to make TIMIT_labels.npy file ?
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/mravanelli/SincNet/issues/12, or mute the thread https://github.com/notifications/unsubscribe-auth/AQGs1tpM1zJmQMkm77ozhrYcuZTMRO7Gks5vGsRJgaJpZM4aScBz .
yeh, why did you use .npy fomat?
Pickle is probably the best option, but it is possible to save/load a dictionary also using Numpy. It has very simple save/loading function (using np.save, np.load).
But yes, you can use any format here. What is really important is to end up with a dictionary with the information reported above.
On Fri, 25 Jan 2019 at 20:07, eeric notifications@github.com wrote:
yeh, why did you use .npy fomat?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/mravanelli/SincNet/issues/12#issuecomment-457785972, or mute the thread https://github.com/notifications/unsubscribe-auth/AQGs1nlB8jvOsVJNYAh0aGZi5DabrlEfks5vG6o4gaJpZM4aScBz .
ok, thank you !
how to make TIMIT_labels.npy file ?