interactiveaudiolab / penn Goto Github PK

#3 demonstrates that stereo audio throws an error. The solution is to convert to mono. A more descriptive error or warning would help.

Nice work, I just read it this morning.

The CLI uses --files, whereas the docs say --audio_files.

(p.s. I'm curious if you have the results tables with for PTDB and MDB-STEM-SYNTH individually? Table II seems to indicate that the scores are mixed.)

It seems both pad=True and pad=False are not zero centered. When Pad=True, the first frame starts from -(winsz-hopsz)//2. instead of -winsz//2.

When using this model for audio at sample rate of 22.05kHz at a hop size of 256, the rounding in the time_to_samples causes the audio hop size to be inaccurate. causing the number of frames to be bigger/smaller than what the hopsize field indicates.

Hi,

I'm working on a project similar to yours, but solely focused on guitar pitch recognition. To have a better look into penn models training I've integrated Weights & Biases into the project, checkout my fork. I'm pretty sure the only thing I've changed in the config is the LOG_INTERVAL value, by setting it to 500, however the training and validation accuracy reported during the training oscillate around 50%, similar results are reported by the evaluation done after the model is trained.

The figures below are the result of my take at training the fcnf0++ model from scratch.

Training accuracy reported every epoch

Validation accuracy reported every epoch

Training loss

The estimated performance is included in the overall.json file generated by the training script.

It's clear that I'm missing something, do you have any advice on steps to achieve the best results, perhaps some issues in my take that are obvious? I tried to follow the README instructions, download, preprocess and partition the mdb and ptdb datasets according to fcnf0++ config, then run the training. In the overal.json file it's reported that evaluation on mdb reaches around 60% and ptdb only around 20%.

overall.json

After a few thousand files of batch inference, the pitch and periodicity become misaligned. This is some sort of multiprocessing issue.

"Cross-domain Neural Pitch and Periodicity Estimation"

Can we export vocals into MIDI? Such as Omnizart Vocals ?

Hi team - can't wait to try this! I'm getting the following RuntimeError when trying to run inference with the pretrained model:

RuntimeError                              Traceback (most recent call last)
[<ipython-input-5-487bd8f6f6cf>](https://localhost:8080/#) in <module>
     29 
     30 # Infer pitch and periodicity
---> 31 pitch, periodicity = penn.from_audio(
     32     audio,
     33     penn.SAMPLE_RATE,

[/usr/local/lib/python3.8/dist-packages/torch/nn/modules/conv.py](https://localhost:8080/#) in _conv_forward(self, input, weight, bias)
    307                             weight, bias, self.stride,
    308                             _single(0), self.dilation, self.groups)
--> 309         return F.conv1d(input, weight, bias, self.stride,
    310                         self.padding, self.dilation, self.groups)
    311 

RuntimeError: Given groups=1, weight of size [256, 1, 32], expected input[2048, 2, 993] to have 1 channels, but got 2 channels instead

I'm getting this with both CPU and GPU inference, and having installed both via pip and having cloned from Github. Do you know what might be the problem?