msaligs/ast_fine_tuned_music_genre_10

🎵 Music Genre Classification using AST

📌 Model Overview

This model is a fine-tuned Audio Spectrogram Transformer (AST) for music genre classification.

It predicts one of the following 10 genres:

• blues, classical, country, disco, hiphop
• jazz, metal, pop, reggae, rock

---

🧠 Architecture

• Base Model: MIT/ast-finetuned-audioset
• Type: Transformer (Audio Spectrogram Transformer)
• Framework: PyTorch + Hugging Face Transformers

---

🎯 Task

Audio Classification (Music Genre Classification)

---

📊 Dataset

• Training Data: Clean instrument stems (drums, vocals, bass, others)
• Test Data: Noisy mashups with:
  • cross-song mixing
  • tempo variation
  • ESC-50 noise injection

---

⚙️ Preprocessing

• Sampling rate: 16kHz
• Fixed duration: 15 seconds
• Padding / truncation applied
• Feature extraction using ASTFeatureExtractor

---

📈 Performance

• Metric: Macro F1 Score
• Achieved: 0.87

---

🔗 Live Demo

Try the model here:
👉 https://huggingface.co/spaces/msaligs/music-genre-classifier

🚀 Usage

from transformers import ASTForAudioClassification, ASTFeatureExtractor

model = ASTForAudioClassification.from_pretrained("msaligs/ast_fine_tuned_music_genre_10")
feature_extractor = ASTFeatureExtractor.from_pretrained("msaligs/ast_fine_tuned_music_genre_10")