Is there a way to add a layer before giving audio as an input to DeepSpeech model???
Could you please elaborate ? You can modify the model, but adding a layer without changing the model that’s not possible.
I want to do some modification in the audio by assuring it will be compatible to give it as input to DeepSpeech.
I want to add this layer1 in DeepSpeech model.
It makes no sense on your diagram. What’s “audio” here ? What’s “Generate audio” ?
layer1 is like pre-processing step of an audio/speech(Audio u can consider it as speech here) which I want to add, generate audio is the audio which is compatible to DeepSpeech input is generated from layer 1.
in simple words, consider DeepSpeech model as X,I have to add a layer whose output is given as input to X.
Yes, you’re free to modify the audio as you see fit. The API takes audio samples as input, so you can modify the audio as much as you want and then just feed it normally.
I want to backpropagate till layer 1 for gradient calculation.How is it possible??Please help me with this issue.
If you know what backpropagation is surely you know enough about TensorFlow to go and edit the code yourself. We’re not here to do your homework.