Sound-Anomaly-Detection

Extract mel spectrogram from wave, train them, test the trained model

Installation

Preprocessing

extract spectrogram

inspection/parameters.py: Modify parameters using in preprocessing.

inspection/model_info.py: Modify classes. These will be used when calculating metrics and training

inspection/preprocess_fns.py: Create your own preprocess method and add its parameters to parameters.json

Training

• tdms files need to be prepared
• file name sould be 'date_label_file name.extension'
  • ex) 210930-GOOD-sample1.tdms
  • '-' is a delimiter. you can change it at consts.py

models/: Trained model will be saved in this directory.

$ python train.py

Test

tdms files need to be prepared

$ python test.py