AI Audio Analysis (WIP)

AI4ALL Class3-Group1

Project Description

Project will showcase an AI model's ability to deconstruct songs, predict their metadata, and then possibly make a genre-based or mood-based playlist for a user to enjoy.

Project Created & Devleloped by:

• Andrew Hernandez (during course)
• William Coleman (currently developing)

Reasearch Question

How accurately can an AI model predict a song's emotional mood based on extracted audio features compared to human perception?

Key Results

• Instrument able to be predicted from any given song's audio features. Looking to increase accuracy and the number of predicted instruments.

Methods

• Semi-supervised Learning

Data Sources

• EmoMusic
• GTZAN
• IRMAS

Technologies Used

Programming Languages

• Python
• Mojo

Frameworks & Libraries

• Keras
• Librosa
• Matplotlib
• NumPy
• Pandas
• Pydub
• Scikit-learn
• Streamlit
• Tensorflow