A CNN model that uses MFCCs from the wav files to train, test & predict the native language of the speaker.
Setup & Requirements • Folder Structure & Code • Running The Project
- Install
Miniconda or Anaconda
- Install
PyCharm
- Open
PyCharm
and hitNew Project
- In the
Create Project
window, select
andNew environment using Conda
Python version: 3.8
- Hit
Create
After creating, the project will open. From the bottom, click on Terminal
. Then:
conda install numba
pip install librosa
pip install matplotlib
pip install --upgrade pip
pip install tensorflow
pip install easygui
- Copy and paste all three
Python
scripts from this repository to thePyCharm
project folder you just created - Copy and paste the folder
Full Dataset wav
from this link to the samePyCharm
project folder
- Use
Alt+Shift+F10
from insidePyCharm
to run each file in the following order - Run
prepare_dataset.py
to preprocess the dataset and create ajson
file with all processed data - Run
cnn_train_test.py
to train the CNN model and save it - Run
predict.py
to load the saved CNN model and use it to classify anywav
file selected using the file explorer