---
title: Audio Classification App
emoji: 🎧
colorFrom: indigo
colorTo: blue
sdk: docker
app_port: 8501
tags:
  - streamlit
pinned: false
---

# 🎧 Audio Classification App

This Streamlit app allows you to upload `.wav` or `.mp3` audio files and get a classification prediction using either an **XGBoost** or **LightGBM** model. It uses pre-trained models and a full pipeline that includes audio preprocessing and feature extraction.

## 🚀 How it Works

1. Upload your audio file (supports `.wav` or `.mp3`).
2. Select a model from the dropdown.
3. The app:
   - Preprocesses and extracts features from the audio.
   - Loads the selected model.
   - Returns the predicted class and inference time.

## 📦 Technologies Used

- `Streamlit` for the frontend
- `librosa` and `soundfile` for audio handling
- `LightGBM` and `XGBoost` for model inference
- `scikit-learn` and `pandas` for data preprocessing

## 🛠 Project Structure

.
├── app.py # Streamlit app
├── requirements.txt # Dependencies
├── README.md # This file
├── models/
│ ├── xgboost.py
│ └── lightgbm.py
├── modules/
│ ├── preprocessing.py
│ └── feature_extraction.py

## 🧪 Inference Pipeline

The app uses a modular inference pipeline built around:

- `AudioPreprocessor`: for loading and cleaning audio signals
- `FeatureExtractor`: for generating MFCC-based feature vectors
- `ModelPipeline`: loads and predicts with the chosen model

## 📄 Requirements

Ensure the following packages are listed in `requirements.txt`:

streamlit
librosa
xgboost
lightgbm
numpy
pandas
scikit-learn
soundfile

## 🔗 Resources

- [Hugging Face Spaces Documentation](https://huggingface.co/docs/hub/spaces)
- [Streamlit Documentation](https://docs.streamlit.io/)