jsleep / wav2mid Goto Github PK

View Code? Open in Web Editor NEW

310.0 310.0 63.0 13.16 MB

Automatic Music Transcription with Deep Neural Networks

License: MIT License

Jupyter Notebook 99.41% Python 0.59%

wav2mid's People

Stargazers

Watchers

Forkers

hpnhxxwn aspriya lwc0329 nindidooo eminarcissus nosignal00 renewu1117 kitchensinkcollection legendfc hallboozhang syncshan aitorbajo yushuyao lngao music-transcription suziw ak9250 zoonono justhalf ankitshah009 metrofun zc280330 samurainote aiedward supershinyeyes yfliao nawalgupta ik0nyshev oisee linyanghao arshad221b questandachievement7developer rippleytrigger jackxavier wwpww meesokim itaybendom leftradio brianpow czf8046 insmoin htbkoo carlosarturome wingong lvyitian lvyitian1 zishanqin thankspei ananya0102 jaedukseo tensor1024 cliffwang2018 tearitco emilie-xu leonas2000 teekhathattt romero8688 bisratgetnet jiahui-zhao2018

wav2mid's Issues

how do i run it

Numpy reshape error while training | ValueError: cannot reshape array of size 1362680 into shape (7,352)

Hello, I am now attempting to train a new model using a new dataset, and the code was running Well until I encountered this issue.
ValueError: cannot reshape array of size 1362680 into shape (7,352).

It was raised in readmm function from File "...AMT Projectswav2midpreprocess.py", line 26, and readmm was called from keras train.py in line 194
[ trainGen = DataGen(os.path.join(path,'data','train'),batch size,args) ]

if anyone can assist.
pls.

how do i use this im confused

im trying to convert my wave file to a midi and i dont know how to do it

Pretty_midi error

Hello, we are trying to run the pre-processing code. We were previously able to run it.
However, suddenly we are getting an error like :

Basically, it seems like a fluidsynth import error.
Could you please suggest how to solve this issue? Thanks.

Memory Error in Preprocessing MAPS dataset

Hi Jon,

We were trying to run the MAPS dataset, with the preprocessing setup as in this repo.
However, we get an error called "memoryerror", in the np. concatenate call in lines -
inputs = np.concatenate(inputs)
outputs = np.concatenate(outputs)

Any suggestions would really help us .

Thanks a lot.

Please upload a trained model

I want to test it on non-piano sound, but my gpu sucks.

question in a comment

in the wav2input function of the preprocess file there's a comment that says:

IMPORTANT NOTE:

Since we pad the the spectrogram frame,

the onset frames are actually `offset` frames.

To obtain a window of the center frame at each true index, we take a slice from i to i+window_size

starting at frame 0 of the padded spectrogram

can you explain it to me, i don't understand it

thank you.

Wrong in Keras_train

In Keras_train.py line 192: DataGen (), and preprocess.py line 39.
There is nothing in the self.mmdirs. Just wondering what is suppose to be in the mmdir?

why i cant't download the dataset

i go to the dataset donload site but it need login in when download dataset, and i do not have the account

Nans in training

Hi Jon,

We are yet facing another issue. We have been able to proceed to the training part.
But the training loss is going to nan when the training data has "AkpnCGDd" data set in it. On removing it the training proceeds fine. When ran just by itself it runs fine. So we are unable to figure out what is causing this issue, if it is this data set or something else.

Any suggestions would be of great help. Thanks.