Sound detection and classification tool based on mel-frequency cepstral coefficients extraction (finger snapping sound used as an example)
1. git clone https://github.com/dustedDuke/CNN-MFCC-sound-detection.git
2. cd CNN-MFCC-sound-detection
3. pip install -r requirements.txt
4. python make_prediction.py <filepath>- train_model.ipynb - covers steps for model training with theory description
- make_prediction.py - standalone prediction tool with matplotlib output
- config.ini - configuration of make_prediction.py
- models - folder containing saved models
- train.csv - train folder description file
- test.csv - test folder description file
- train - folder containing train dataset (can be downloaded from https://zenodo.org/record/2552860/files/FSDKaggle2018.audio_train.zip?download=1)
- test folder containing test dataset (can be downloaded from https://zenodo.org/record/2552860/files/FSDKaggle2018.audio_test.zip?download=1