Gherman Sebastian-Costin Politehnica University of Bucharest, CS, Deep Learning Algorithm to identify car engine types
Based on the following article:
Spectral features for audio based vehicle and engine classification
Alicja Wieczorkowska, Elżbieta Kubera, Tomasz Słowik & Krzysztof Skrzypiec
Journal of Intelligent Information Systems
The majority of the code is based on Paul Mora's approach to determining
differences between two Formula 1 engine manufacturers
https://becominghuman.ai/signal-processing-engine-sound-detection-a88a8fa48344
Code improvements and completion based on Valerio Velardo's Deep Learning (for Audio)
with Python course on Youtube (great course)
https://www.youtube.com/playlist?list=PL-wATfeyAMNrtbkCNsLcpoAyBBRJZVlnf
Massive thanks to stackoverflow.com and https://www.tensorflow.org/
Usage:
python3 getter.py argv {> output}
argv :
1 - only examples to run on 2 chosen diesel-gasoline input files;
Outputs the following:
/output/mfccs.png
/output/powerspectrum.png
/output/short_fourier.png
/output/waveplot.png
2 - train algorithm
Firstly it creates the dataset in /data/diesel and /data/diesel in .wav format
data must be provided in /raw/diesel and /raw/gasoline in .wav format
Outputs the following:
output/accuracy_error.png
data/processed_data.json
/code/my_model - as it saves the trained model for future use
3 - test algorithm
Must have in /raw/diesel and /raw/gasoline test files with the name test{number}.wav
Outputs the following:
code/predictions_output
L.E Uploaded requirements.txt