ARIA

AIRA is an AI‑powered piano mentor built by Team C242‑PS540 for the Bangkit Academy 2024 Capstone. A real-time piano learning assistant that detects note and rhythm errors in under 300 ms, providing instant feedback to students via an Android client. Built with a Flask backend for low-latency audio processing and machine learning classification.

• Real‑Time Feedback: Classifies Wrong Note, Missing Note, Extra Note, or No Error against reference MIDI with ~92% accuracy.
• Low-Latency Feedback – End-to-end processing time under 300 ms.
• Cross-Platform Streaming – Streams audio from Android app to Python backend.
• ML-Driven Classification – Trained on annotated piano audio datasets for robust detection.
• Immediate Learning Aid – Students can self-correct mistakes mid-practice session.

Tech Stack

• Backend: Python (Flask)
• ML Libraries: TensorFlow, Keras, Librosa
• Client: Android (Java/Kotlin)
• Other Tools: NumPy, Pandas, GitHub Actions for CI/CD

Screenshots

Practice Mode

Error Feedback Detail

Installation

Clone and set up each component:

Android App

git clone https://github.com/TCHWG/Android‑Development.git
cd Android‑Development
./gradlew assembleDebug

Backend & API

git clone https://github.com/TCHWG/Backend‑AIRA.git
cd Backend‑AIRA
npm install
cp .env.example .env   # fill in your Firebase, email, and Google Cloud creds
npx prisma migrate dev
npm start

Machine Learning Model

git clone https://github.com/TCHWG/Machine‑Learning.git
cd Machine‑Learning
pip install -r requirements.txt
python train_model.py   # trains the mistake‑classification model

Usage

1. Launch the Android app and sign up
2. Connect to the backend URL in Settings
3. Start a new practice session; AIRA listens via mic/MIDI
4. Review instant feedback overlays and badges earned
5. Check your weekly progress on the Dashboard screen