Join GitHub today
GitHub is home to over 40 million developers working together to host and review code, manage projects, and build software together.Sign up
Fix display (SG and Audio, show_batch) #15
Fix the AudioSpectrogram and AudioItem display to be proper and compatible with how fastai handles display.
It is tricky because we aim to support multichannel audio, and multiple images per channel (e.g. MFCC + it's delta/accelerate for each audio channel). This means a single stereo audio clip that has MFCC + delta/accelerate will be 6 images, this creates a problem for stuff like show_batch and I don't know how to resolve.
We also want to have a grid + colorbar which I've had difficulty with for multiplot display.
I propose someone more familiar with matplotlib and fastai display rewrite it to be fully functional
I'd also like to add a parameter to AudioSpectrogram's to show a audio player beneath them. It might not be too hard and will be useful to hear your spectrograms! After awhile it you can get lost as to what you are looking at sounds like :)
This should also work properly as expected for show_batch() for training purposes
and change it to:
its showing batches.
Does this mean that everything is ok? What is the expected/ideal behavior? If there is anymore information you can provide, I can keep working on this! Let me know.
Here is the branch where I'll be working on this: https://github.com/aquietlife/fastai2_audio/tree/fix_display