The necessary packages to run the code included in this project are listed in requirements.txt.
The best option to install these packages is to create a new virtual environment with Python 3.11
and install the packages within that environment.
pip install -r requirements.txt
Before running this code, you need to have a folder named chillanto outside this directory.
The Chillanto folder must contain all the crying samples. The main classifier is programmed to classify
5 classes, but this depends on the folders. If there are only 3 folders, then there will be 3 classes.
If more or fewer classes are needed, organize the folders by class and delete the chillanto_metadata.csv
file to generate a new one. The model I trained has these 5 classes:
- asphyxia
- deaf
- normal
- hunger
- pain
Inside this folder, there should be other folders containing the crying samples, separated by class in each folder.
I have included the chillanto_metadata.csv file to clarify any doubts regarding the folder organization.
Once the samples are ready, simply run the main.py script:
python main.pyAfter the model finishes training, two versions of the model will be saved in the saved_models
folder; one in hdf5 format and the other as json and h5 files. To load the trained model and test it to
generate a confusion matrix, run the test_model.py script:
python test_model.py