drscotthawley / audio-classifier-keras-cnn

Audio Classifier in Keras using Convolutional Neural Network

audio-classifier-keras-cnn

Audio Classifier in Keras using Convolutional Neural Network

DISCLAIMER: This code is not being maintained. Your Issues will be ignored. For up-to-date code, switch over to Panotti.

audio-classifier-keras-cnn's Issues

Train_network.py can't handle smaller mel spectrogram shapes

Hi Dr. Hawley,

I noticed a small problem in the code in both train_network and eval_network- there is no error handling for files that produce spectrograms smaller than width 1293. This happens leads when the training data is created from the mel spectrograms (X_train[train_count,:,:] = melgram, around line 140).
You have written code to chop off the extra width if it is too long ( melgram = melgram[:,:,:,0:mel_dims[3]] ) but nothing to account for melgrams being too short.
I was able to get around it by filling the empty space with 0's, but I thought it would be helpful to let you know!

Also- if you are interested, I would love to connect with you sometime to talk about potential ways to extend this example/model to a system that works in real time, and makes predictions on songs as it hears them through a computer microphone versus an uploaded mp3.
My email is [email protected] if you want to connect!

Thanks,

Aaron

ValueError: could not broadcast input array from shape (1,96,25) into shape (1,96,32)

Dear Professor, Thank you for the excellent tutorial.

For sample data, i created two folders and put one set of sounds in each folder.

Step 1 went well.

man@man:~/Downloads/audio-classifier-keras-cnn-master$ python preprocess_data.py 
class_names =  ['tring', 'hold']
 class name = tring          -   0, 9 files in this class
 Loading class: tring          ( 1 of  2 classes), file 1 of 9: Samples/tring/r4.wav
 class name = hold           -   1, 94 files in this class
 Loading class: hold           ( 2 of  2 classes), file 1 of 94: Samples/hold/hm1_081.mp3
 Loading class: hold           ( 2 of  2 classes), file 21 of 94: Samples/hold/hm1_011.mp3
 Loading class: hold           ( 2 of  2 classes), file 41 of 94: Samples/hold/hm1_020.mp3
 Loading class: hold           ( 2 of  2 classes), file 61 of 94: Samples/hold/hm1_077.mp3
 Loading class: hold           ( 2 of  2 classes), file 81 of 94: Samples/hold/hm1_043.mp3

STEP TWO ERROR

man@man:~/Downloads/audio-classifier-keras-cnn-master$ python train_network.py
Using TensorFlow backend.
class_names =  ['tring', 'hold']
total files =  103
   get_sample_dimensions: melgram.shape =  (1, 1, 96, 32)

 Loading class: tring          ( 1 of  2 classes), file 1 of 9: Preproc/tring/r6.wav.npy
Traceback (most recent call last):
  File "train_network.py", line 201, in <module>
    X_train, Y_train, paths_train, X_test, Y_test, paths_test, class_names, sr = build_datasets(preproc=True)
  File "train_network.py", line 147, in build_datasets
    X_train[train_count,:,:] = melgram
ValueError: could not broadcast input array from shape (1,96,25) into shape (1,96,32)
man@man:~/Downloads/audio-classifier-keras-cnn-master$ 

What is meant by ValueError: could not broadcast input array from shape (1,96,25) into shape (1,96,32) How do I resolve it?