Hi Valerio! Can u demonstrate doing this on how to implement data augmentation to all wav files inside the folder, instead of augmented one by one file?
regarding the 'weirdness' of time scaling, you mention that it shouldn't be "abused" but isn't that the point of a ML based approach? if the dataset is trained on what a piano or drum kit or whatever is supposed to sound like shouldn't it be able to account for these types of tempo/pitch changes and avoid distortion?
just in case anyone is having trouble with keyword arguments, you must specify parameters as keyword arguments either in the function definition of the librosa call or else it will not run properly
TypeError: pitch_shift() takes 1 positional argument but 2 positional arguments (and 1 keyword-only argument) were given I have this error, can you please help
Hi Valerio ! Lately I am working on an imbalanced speech dataset and I would like to ask if data augmentation is the only way to increase number of the minority class records. Do you think a combination of augmenting the minority classes and undersampling the majority classes would help?
let me have your idea on the following ideas how can I extract texture feature from audio spectrograms and which one is best for almost 1 second audio , and does Vq (vector quantization be used) for audio classification with python ?
thanks for all series that you've shared to everyone
You're welcome!
The serie we didn't deserve but we needed ahhah
hi Valerio, I have a question..how to implement augmented to all wav files inside the folder, instead of augmented one by one file?
Amazing introduction. Learned a lot! Thanks for sharing!
Nice green video) augmentation in action
Hi Valerio! Can u demonstrate doing this on how to implement data augmentation to all wav files inside the folder, instead of augmented one by one file?
Thanks a lot. You are an excellent instructor.
Thanks a lot Valerio, you're doing excellent work, and great explanation.
Thanks!
regarding the 'weirdness' of time scaling, you mention that it shouldn't be "abused" but isn't that the point of a ML based approach? if the dataset is trained on what a piano or drum kit or whatever is supposed to sound like shouldn't it be able to account for these types of tempo/pitch changes and avoid distortion?
Can we do the same for vibration data augmentation?
i have a tuple error with signal.std() Please help. 'tuple' object has no attribute 'std'
Thanks buddy! You helped me a lot :)
just in case anyone is having trouble with keyword arguments, you must specify parameters as keyword arguments either in the function definition of the librosa call or else it will not run properly
TypeError: pitch_shift() takes 1 positional argument but 2 positional arguments (and 1 keyword-only argument) were given
I have this error, can you please help
Hi Valerio ! Lately I am working on an imbalanced speech dataset and I would like to ask if data augmentation is the only way to increase number of the minority class records. Do you think a combination of augmenting the minority classes and undersampling the majority classes would help?
Tryout loss weights
let me have your idea on the following ideas
how can I extract texture feature from audio spectrograms and which one is best for almost 1 second audio , and does Vq (vector quantization be used) for audio classification with python ?
nice explaination
You and your videos are alsom, thanks :)