https://github.com/gastonduault/music-classifier

This project uses AI for analyzing songs. With the customized and trained model and the website with Flask you can classify music genres.
https://github.com/gastonduault/music-classifier

flask huggingface model music torch transformer wav

Last synced: about 1 month ago
JSON representation

This project uses AI for analyzing songs. With the customized and trained model and the website with Flask you can classify music genres.

• Host: GitHub
• URL: https://github.com/gastonduault/music-classifier
• Owner: gastonduault
• Created: 2024-11-18T09:53:03.000Z (about 2 months ago)
• Default Branch: main
• Last Pushed: 2024-11-18T12:59:03.000Z (about 2 months ago)
• Last Synced: 2024-11-18T14:32:17.479Z (about 2 months ago)
• Topics: flask, huggingface, model, music, torch, transformer, wav
• Language: Python
• Homepage:
• Size: 63.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Music Genre Classification with AI🎶

This **[model](https://huggingface.co/gastonduault/music-classifier)** classifies music genres based on audio signals. It was fine-tuned on the

**[music_genres_small](https://huggingface.co/datasets/lewtun/music_genres_small)** dataset.

This project leverages the **[Wav2Vec2](https://huggingface.co/facebook/wav2vec2-large)** model from Facebook, a state-of-the-art transformer-based architecture for audio processing, to classify music genres. The model is fine-tuned on a dataset of music samples to predict the genre of a given audio file.

The repository includes:

- Preprocessing of audio data.

- Flask service for a website for try the model

- Training scripts for fine-tuning the Wav2Vec2 model.

- Model evaluation and performance analysis.

- Deployment of the trained model for real-world predictions.

---

## Getting Started

### Prerequisites

- Python 3

- `pip install transformers werkzeug flask torch librosa`

### Run API Flask

- `Python app.py` (run app.py)

- website on **[http://localhost:5000](http://localhost:5000)**

---

## 🖥️ Access the Model

➡️ **[Fine-Tuned Model gastonduault/music-classifier](https://huggingface.co/gastonduault/music-classifier)**

### Metrics

- **Validation Accuracy**: 75%

- **F1 Score**: 74%

- **Validation Loss**: 0.77

---

## 📊 Dataset

The dataset used for training is `lewtun/music_genres_small` from Hugging Face's datasets library. It contains diverse audio tracks labeled with genres.

**Classes:**

- `Electronic`

- `Rock`

- `Punk`

- `Experimental`

- `Hip-Hop`

- `Folk`

- `Chiptune / Glitch`

- `Instrumental`

- `Pop`

- `International`

---

## 📂 Repository Structure

- **`training.py`**: Script for training and evaluating the Wav2Vec2 model ***(with Google colab)***.

- **`prediction.py`**: Script to make predictions on new audio files using the trained model.

- **`app.py`**: Python API with Flask framework for testing the model.

---

## The interface of testing

![img.png](static/img.png)