search

ZahraZali / NoiseCut

GNU Affero General Public License v3.0
20 stars 4 forks source link

Unable to install #2

Open SquirrelKnight opened 3 months ago

SquirrelKnight commented 3 months ago

Hello, I am currently having issues installing NoiseCut. So far, I have attempted to create a new conda environement "conda create -n noisecut python=3.9". After activating the environment, I then run "python setup.py install" from the NoiseCut directory, but the installation fails at the scipy step. If I manually install scipy, it then fails at the obspy step. Once again, if I manually install obspy, there is an incompatibility between the version of matplotlib and numpy.

Do you have any suggestions to help me get this working? I am working with OBS data, and this seems like a very promising method! Thanks for your time.

SquirrelKnight commented 3 months ago

To update, I downgraded numba to version 0.57.1. This seems to have resolved the issue. I also installed using pip.

From the NoiseCut directory I just ran 'pip install .'

As a couple of additional questions. Is there a minimum waveform time window? For example, I had a one hour waveform, but NoiseCut failed to execute... however it worked well for a one day window.

Does the sample rate make a difference? If I have > 100 Hz data, should I downsample? I have some seismometers with 500 Hz channels.

Thanks!

ZahraZali commented 3 days ago

Hello,

Thank you for your interest in NoiseCut, and I apologize for the delayed response. I'm glad to hear that you were able to install the package successfully.

To answer your questions:

Minimum Waveform Time Window: If you’re using the automated package (by importing NoiseCut), it may not work as effectively on shorter waveforms. In your case, the one-hour waveform might be too short for the current settings, which is likely why it failed. A one-day window is generally recommended for optimal performance. However, you can modify the STFT parameters in the source code to better suit your waveform length and run the code directly (without importing NoiseCut). I'll work on updating the package to be more flexible with different waveform sizes.

Sample Rate: Sample rate and waveform window are interrelated. The package works well for one-day data with a 100 Hz sampling rate. If your data has a 500 Hz sampling rate, you might want to use 4 to 6 hours of data to get the best performance with the default settings. Again, you can adjust the parameters in the source code to better fit your data.

Let me know how your progress goes, and feel free to reach out if you have any further questions.