Mel Spectrogram CNN

Project Overview

This machine learning project demonstrates deep learning expertise by building a Convolutional Neural Network (CNN) to classify music by artist using mel spectrogram audio representations. The project showcases a complete ML pipeline from data processing to model training, converting raw audio files into visual representations optimized for neural network analysis.

Related ML Projects

This project is part of a broader portfolio of machine learning applications, including:

  • Student Stress Prediction: Multi-model ML system using Neural Networks, Random Forests, and Support Vector Machines
  • Lung Disease CNN: Deep learning classification on 8000+ medical images from Kaggle
  • YouTube Video Summary AI: NLP application for automated video content summarization
  • DQN Reinforcement Learning: PyTorch-based reinforcement learning for CartPole control problem

Key Features

  • Audio Processing: Converts MP3 audio files to mel spectrograms using librosa library
  • Data Pipeline: Implements automated data loading and preprocessing from multi-level directory structure
  • Data Visualization: Uses Plotly for interactive visualization of audio features
  • Deep Learning: Builds CNN model using Keras/TensorFlow for artist classification
  • Scientific Computing: Utilizes NumPy and Pandas for data manipulation and analysis

Technology Stack

Python TensorFlow Keras NumPy Pandas Librosa Matplotlib Plotly

What's Included

  • Complete Jupyter notebook with documented code
  • Data loading and preprocessing pipeline
  • Audio feature extraction and visualization
  • CNN model architecture and training
  • Evaluation metrics and performance analysis

Skills Demonstrated

  • Machine Learning & Deep Learning
  • Audio Signal Processing
  • Data Science Workflow
  • Python Programming
  • Data Visualization
  • Neural Network Architecture Design