encoradigital / sab-cnn-audio-denoiser Goto Github PK

Code is giving error in create_dataset script val_dataset.create_tf_record(prefix='val', subset_size=2000) while creating subsets of training and validation dataset

Hello. Is it possible that you can share the pretrained model or in other words can you share the weights of the models that can speed up the things.

It would really be great to have the pertained model in .h5 file in order to use in transfer learning in other noise reduction projects. Thank you very much for your help!

When running Pathon code about !wget 'cdn.daitan.com/dataset.zip', there is no data placed in the website. Could you upload data again. Thanks.

Code is giving error in val_dataset.create_tf_record(prefix='val', subset_size=2000).

Error:
Traceback (most recent call last):
File "create_dataset.py", line 33, in
val_dataset.create_tf_record(prefix='val', subset_size=2000)
File "/content/data_processing/dataset.py", line 140, in create_tf_record
writer = tf.io.TFRecordWriter(tfrecord_filename)
File "/usr/local/lib/python3.6/dist-packages/tensorflow/python/lib/io/tf_record.py", line 299, in init
compat.as_bytes(path), options._as_record_writer_options())
tensorflow.python.framework.errors_impl.NotFoundError: ./records/val_0.tfrecords; No such file or directory

Hii, I want to denoise a noisy signal in an unsupervised way.can this be possible using this model.

After running model, it does not converge and remove noise. Any one got it working

您好，数据集的链接失效了，可以重新提供一下吗？

This project is broken !
Can you get this working with proper instructions please, otherwise you are wasting peoples time even reading the medium article !

And its a very bad advertisement for your company !!!!

a should be l2_norm(clean) in your SDR function.

def SDR(denoised, cleaned, eps=1e-7): # Signal to Distortion Ratio
a = l2_norm(denoised) <--- here
b = l2_norm(denoised - cleaned)
a_b = a / b
return np.mean(10 * np.log10(a_b + eps))

E. Vincent, R. Gribonval, and C. F´evotte, “Performance measurement
in blind audio source separation,” IEEE transactions on audio,
speech, and language processing, vol. 14, no. 4, pp. 1462–
1469, 2006.

I cannot download the original dataset from the link that has been provided here as I am getting 404'd when going to the link that has provided and I've been getting an error when running "create_dataset.py" with python 3.8.

File "create_dataset.py", line 24, in
val_dataset.create_tf_record(prefix='value', subset_size=2000)
File "/home/gabriel/cnn-audio-denoiser/data_processing/dataset.py", line 140, in create_tf_record
writer = tf.io.TFRecordWriter(tfrecord_filename)
File "/home/gabriel/anaconda3/lib/python3.8/site-packages/tensorflow/python/lib/io/tf_record.py", line 298, in init
super(TFRecordWriter, self).init(
tensorflow.python.framework.errors_impl.NotFoundError: ./records/value_0.tfrecords; No such file or directory

If this isn't enough information, please tell me, as I'd be happy to help.

Sincerely,
Gabriel

Edit: creating a folder called records on cnn-denoiser fixed it.

ValueError: Output tensors to a Model must be the output of a TensorFlow Layer (thus holding past layer metadata). Found: Tensor("conv2d_159/BiasAdd:0", shape=(?, 129, 1, 1), dtype=float32)

I have tried this suggestion , It was not working even after this. But I am not sure where to wrap with lambda function.
"https://stackoverflow.com/questions/50715928/valueerror-output-tensors-to-a-model-must-be-the-output-of-a-tensorflow-layer"

Please Help !!