search

Jakobovski / free-spoken-digit-dataset

A free audio dataset of spoken digits. An audio version of MNIST.
626 stars 248 forks source link

Save spectrograms as desired without post-processing and python3 compatibility #24

Closed felixdollack closed 5 years ago

felixdollack commented 5 years ago

Dear @Jakobovski, I tried to use this dataset (awesome idea and work by the way) with python 3 and had some trouble with the printing calls. I imported the print_function from the future module to keep it python 2 compatible. I also saw the todo comment in the spectrogram script and changed the plot command accordingly to create figures without padding and resizing as desired. During testing I fixed a spelling mistake with the colormap that triggers a ValueError with matplotlib. If you like it and think it is useful I would be happy if you merge the changes :)

Best regards, Felix

Jakobovski commented 5 years ago

Looks great. Thanks!

Jakobovski commented 5 years ago

Out of curiosity what are you doing with the dataset?

felixdollack commented 5 years ago

I am relatively new to the machine learning train, but I always wanted to try something similar to your decoupled-multimodal-learning repository. But then I had trouble with the print statements and started with this dataset. I might push something similar for the other project soon ;)