19. 🎤 Music Genre Classifier

Overview

An audio tool that labels uploaded music tracks with one or more genres, handling both studio recordings and live/jam sessions.

Primary Use Cases

• Music librarians organizing large audio collections
• Playlist curators auto-generating genre-based lists
• Streaming services improving recommendation metadata

Key Features

• Audio upload (MP3, WAV) or URL import
• Preprocessing: trim silence, compute spectrograms via Librosa
• Classification: multi-label CNN or Transformer-based model
• Batch mode: process entire directories
• Results UI: confidence scores, top-3 genre suggestions

Tech Stack

• Frontend: React + TypeScript + Tailwind (upload & dashboard)
• Backend: FastAPI (Python) + Go worker for batch jobs
• AI Models:
  • Feature extraction: Mel-spectrograms via Librosa
  • Classifier: fine-tuned panns_cnn6 or ast-model from Hugging Face

Architecture

1. Audio Service: receives file → normalizes sample rate → generates spectrograms.
2. Inference Service: loads model → returns genre probabilities.