🎵 AudioCompass

An ML-powered audio analyzer with 98% accuracy that helps you understand audio at a glance.

---

📋 Objective

To develop a machine learning–powered audio analyzer that can accurately:

• 🎼 Detect the musical key & scale (e.g., C Major, A♯ Minor)
• 🎵 Estimate the tempo (BPM) of a track

…all from a single audio file using a *shared feature learning model, deployed as an interactive *Streamlit application.

---

🧩 Problem Formulation

• Input:
  An audio file (e.g., .wav, .mp3)

• Output:

  • Key: One of 12 tonal keys (C, C#, D, …, B)
  • Scale: Major / Minor
  • BPM: A continuous numerical value (regression)

---

🚀 Features

✅ Accepts .wav or .mp3 audio input
✅ Predicts:

1. 🎼 Musical key & scale
2. 🎵 Tempo (BPM)

---

📊 Results & Visualizations

Here are some example results from the model evaluation and testing:

🔷 Graph 1

🔷 Graph 2

🔷 Graph 3

🔷 Graph 4

---

📂 How to Run Locally

1️⃣ Clone the repository and navigate to the project directory

git clone https://github.com/mayan05/AudioCompass.git
cd AudioCompass

2️⃣ Install dependencies

pip install -r requirements.txt
cd stream

3️⃣ Run the Streamlit app

streamlit run app.py

4️⃣ Run the FastAPI backend server

uvicorn stream.server:app --reload --port 8000

5️⃣ Upload an audio file and view predictions.

---

🛠 Technologies Used

PyTorch🔥

Streamlit 📈

Librosa 🎧

Scikit-learn 🔬

FastAPI 🍃

Numpy 🔢

---

✨ Contributors

• Mayan Sequeira

  • GitHub: https://github.com/mayan05
  • Email: mayan.sequeira@gmail.com

• Rahul Anand

  • GitHub: https://github.com/RahulAnand2077
  • Email: rahulcollege27@gmail.com