search

apple / turicreate

Turi Create simplifies the development of custom machine learning models.
BSD 3-Clause "New" or "Revised" License
11.2k stars 1.14k forks source link

Sound Classifier Data Augmentation #1571

Open TobyRoseman opened 5 years ago

TobyRoseman commented 5 years ago

The sound classifier should do data augmentation by default and give the users control over how it is done.

Adding pink noise and reverb are probably the two most important augmentation methods. Adjusting the volume levels would also be useful.

If we are doing data augmentation, we probably should not be doing channel averaging at training. Instead we should treat each channel as a separate example.

TobyRoseman commented 5 years ago

@ylow also had the idea of using randomly placed windows.