Project Description:

The AI-Based Music Genre Classification system is designed to automatically detect and classify the genre of a given music file (MP3/WAV) using machine learning and audio signal processing techniques. By extracting meaningful features like tempo, rhythm, pitch, and frequency patterns, the system identifies genres such as rock, jazz, classical, hip-hop, pop, or blues. It can be used in music streaming apps, personal music libraries, or recommendation systems.

Core Objective:

To build a system that uses AI/ML to classify music tracks into their respective genres by analyzing the audio content, not just metadata or file tags.

Key Features:

1. Audio File Upload & Playback

   • Users can upload audio files (MP3, WAV).

   • Preview/play option for uploaded files.

2. Audio Feature Extraction

   • Mel-frequency cepstral coefficients (MFCCs)

   • Chroma features, spectral contrast, tempo

   • Zero-crossing rate, RMS energy

3. AI/ML Genre Classification

   • Trained model using supervised learning (SVM, Random Forest, CNN, or LSTM)

   • Outputs predicted genre with confidence score

4. Result Visualization

   • Bar chart or pie chart showing prediction probabilities

   • Option to compare with top 3 genre matches

5. Dataset Support

   • Pre-trained model on GTZAN or FMA (Free Music Archive) dataset

Tech Stack:

 Audio Processing:

• Librosa (Python) for feature extraction

• FFmpeg for audio format conversion

 Machine Learning:

• Python + Scikit-learn for classical ML (SVM, RF, KNN)

• TensorFlow/Keras for deep learning models (CNN for spectrogram classification)

 Frontend:

• HTML, CSS, Bootstrap

• JavaScript + audio visualizer (optional)

 Backend:

• Node.js / PHP / Java backend for file upload and communication with ML model via API

• Python server (Flask/Django) for ML model inference (optional if needed)

 Database:

• MongoDB or MySQL to store user uploads and prediction history

How It Works:

1. User uploads a music file.

2. System extracts audio features using Librosa.

3. Features are fed into a trained AI model.

4. Model predicts the genre and sends the result to frontend.

5. Visualization and feedback are shown to the user.