jim-schwoebel / sound_event_detection Goto Github PK

label_files.py

Install dependencies. Created new environment and pip installed each one. Requirements.txt didn’t work for some reason. My guess is it hung up when it reached the couple that needed special installation instructions: PyAudio and PocketSphinx. PyAudio was installed with following link: https://stackoverflow.com/questions/54998028/how-do-i-install-pyaudio-on-python-3-7. Jim provided following link to get PocketSphinx going on Windows, but it appears to work fine without it https://stackoverflow.com/questions/18889268/setting-up-pocketsphinx-for-python-in-windows.
Line 205 in ‘label_files.py’: os.system('play %s'%(filename)) doesn’t appear to work with Windows. Command prompt returns some play error. Replaced with below to use in windows. File will then play in command prompt and annotation can take place.
import winsound #(should be native in Windows Python install)
print(filename)
winsound.PlaySound(filename, winsound.SND_FILENAME)
train_audioTPOT.py

Had issue with librosa_features.py: Error says librosa_feature.rmse(y)[0] (line 138) has no attribute rmse. Looing online, it says attribute changed to ‘rms’ instead of’ rmse’. Works after changing to ‘rms’.
When doing quick testing ran into some json file issues. Not exactly sure what happened, but during this test there were only 1 or 2 files in class target 2. I reran and tried to make even number for each class and it works fine.