instruments classification

The objective of this project is to develop a real-time musical instrument recognition system that will accurately identify the different instruments present in the sound signal. To achieve this, we designed a feature extraction system and built a labeled database containing 0.1 s audio sequences. We then trained and compared two convolutional neural network (CNN) and long-term memory (LSTM) models for audio classification. Finally, we developed a Streamlit application for real-time audio analysis.

Dataset:

Freesound plus de 370 000 avec plus de 40 labels https://www.kaggle.com/c/freesound-audio-tagging

For use this repository, you need to install this requirements

python3.6

IPython

Numpy

python_speech_features

h5py

Pandas

Librosa

Scikit-learn

scipy.io

os

keras

Matplotlib

Result

convolutional neural network

Long Short Term Memory