A research project in which a model was trained to detect whether a given audio file contains aggressiveness. This project was extended to also detect whether a given audio file contains an instance of bullying.
The Jupyter notebook can be used to train a model to differentiate between various classes of audio files. To do so, download this project and populate the data folder with pre-sorted .wav audio files. Afterwards, run the notebook to produce .h5, .json, and .yaml files containing the model.
This project also contains a file (flask.py) that can be used to host an acoustic AI algorithm. A sample endpoint that uses this code can be found at https://ericthestein.pythonanywhere.com. To invoke this, send a POST request with a body that contains a form, in which the key, "recording" contains a .wav file.
An example usage of the sample endpoint:
- TensorFlow - a machine learning platform developed by Google
- Keras - a neural network API
- Flask - a web application framework
- Eric Stein
This project is licensed under the GNU License - see the LICENSE.md file for details
- Dr. Anthony Joseph, Pace University - Oversaw the experimental design and progress of this project
- Manash Kumar Mandal - Author of an acoustic deep learning tutorial: https://medium.com/manash-en-blog/building-a-dead-simple-word-recognition-engine-using-convnet-in-keras-25e72c19c12b