kapre
kapre copied to clipboard
trainable DSP parameters
hello contributers and community.
I love your repo! It's eases so much for me! Although having the precomputation in the model is already great I'd like to know how you can optimize DSP parameters. It looks like that this is a feature from old versions (e.g. 0.2) and by default I dont see any trainable params in this layer.
Could you please state if this is still available and how to use it?
happy hacking Paul
Hi, sorry for a late reply. In a preference of using tf.signal.stft
, I removed it in the newer versions of Kapre
.
To be honest, I miss the feature ever since though. But recently I'm mainly using Pytorch so somehow Kapre's having another winter.
Are differentiable bandpass filters (like scipy's butter + sosfiltfilt) possible in keras/tf?
Edit: Looks like this was done with SincNet in Pytorch and adapted to Keras here!