Audio Classification

Datasets

• Speech Commands from torchaudio, with 35 class.

Model Architecture

• Using AlexNet with Mel Spectrogram 1 channel.
• Output is a softmax with 35 nodes (35 class).

How to use this respository?

• Edit your configuration in conf/configs.yaml
• Train model using python main.py -cp conf -cn configs

Additional

• You guys should install PySoundFile on windows or sox on linux, for torchaudio I/O backend.