search

MikeMpapa / CNNs-Speech-Music-Discrimination

A deep learning framework for Speech-Music discrimination of continuous audio streams
MIT License
68 stars 13 forks source link

Increasing number of classes #4

Closed aitorbajo closed 6 years ago

aitorbajo commented 6 years ago

Dear Mike

I am trying to train a 3-class based (music, speech and noise) audio classifier system using your tool.

Do I have to change the given proto somehow?

I trained a 2-classes (music, speech) model days ago and works fine, but when I increase the number of classes and do training, the output model does not work correctly even the training process has finished without errors. The classification gives always the same class as output (music in this case, the first one) without probabilities.

Thank you so much in advance! ;-)

MikeMpapa commented 6 years ago

Hi ,

Thanks for your interest in my work.

To change the number of target classes just open the related .prototxt file and search for the " num_output:" parameter at the last Fully-Connected Layer. Currently is set up to 2 (Speech VS Music). Change that to your preference.

Please "Star" my work if you found it useful.

Thank you, Michalis

Michalis Papakostas

PhD Candidate Heracleia - Human Centered Computing Lab Department of Computer Science University of Texas, Arlington

2018-02-14 12:20 GMT-06:00 aitorbajo notifications@github.com:

Dear Mike

I am trying to train a 3-class based (music, speech and noise) audio classifier system using your tool.

Do I have to change the given proto somehow?

I trained a 2-classes (music, speech) model days ago and works fine, but when I increase the number of classes and do training, the output model does not work correctly even the training process has finished without errors. The classification gives always the same class as output (music in this case, the first one) without probabilities.

Thank you so much in advance! ;-)

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/MikeMpapa/CNNs-Speech-Music-Discrimination/issues/4, or mute the thread https://github.com/notifications/unsubscribe-auth/AMB79tdgcRy92T9gonBeQRL23HaDjFEKks5tUyPvgaJpZM4SFwZ4 .