MET – Metabolic Equivalent of Task AI Android App

An Android application and machine learning pipeline for real-time physical activity recognition.
The app predicts a user’s current MET class (Metabolic Equivalent of Task) from smartphone accelerometer and gyroscope data, and shows cumulative daily activity across:

• Sedentary (< 1.5 METs)
• Light (1.5–3 METs)
• Moderate (3–6 METs)
• Vigorous (> 6 METs)

---

📱 Android App

• Source code lives in: mobile_app/
• Built in Android Studio (Kotlin).
• Uses foreground service for continuous sensing.
• On-device inference with ONNX Runtime.
• Room database stores daily history.
• Live UI: current MET class + pie chart + history view.

To build the APK:

1. Open mobile_app/ in Android Studio.
2. Go to Build → Build Bundle(s)/APK(s) → Build APK(s).
3. The APK will appear in mobile_app/app/build/outputs/apk/release/.

---

🧠 Machine Learning Model

• Hybrid model:
  • 1D CNN → raw windows [150 × 8] (3s @ 50Hz).
  • MLP → 36 engineered features (mean, std, min, max, jerk).
• Datasets: WISDM, MotionSense, UCI HAR.
• Augmentations: random rotation, time warping.
• Exported to ONNX (see Models/onnx/).

---

📊 Evaluation

Results from 5 independent runs on a 20% test set are stored in evaluation_results.txt.

Averaged Results:

Metric	Mean ± Std
Accuracy	95.68% ± 0.90
Precision	95.77% ± 0.77
Recall	95.68% ± 0.90
F1-score	95.68% ± 0.87

---

🚀 Training Pipeline

• Code: train_eval_met_pipeline.py
• Dependencies: requirements.txt

Install dependencies

pip install -r requirements.txt

Run training / evaluation

# Evaluation mode: 5 runs, averages metrics, saves results
EVAL_MODE=True python train_eval_met_pipeline.py

# Normal training mode: trains once and saves ONNX model
EVAL_MODE=False python train_eval_met_pipeline.py

Models and scalers are saved into Models/.

---

📂 Repository Structure

.
├── .vscode/                     # VSCode settings (optional)
├── Models/                      # Trained models, scalers, ONNX
├── data/                        # Placeholder for datasets (WISDM, MotionSense, UCI HAR)
├── mobile_app/                  # Android Studio project
├── evaluation_results.txt       # Evaluation metrics (5 runs averaged)
├── requirements.txt             # Python dependencies
├── train_eval_met_pipeline.py   # Training + evaluation pipeline
└── .gitignore                   # Ignore rules

---

📄 Deliverables

• ✅ Android app source (mobile_app/)
• ✅ Trained ONNX model + scalers (Models/)
• ✅ Training pipeline (train_eval_met_pipeline.py)
• ✅ Evaluation results (evaluation_results.txt)

---

📚 Datasets' References

• Kwapisz, J. R., Weiss, G. M., & Moore, S. A. (2010). Activity Recognition using Cell Phone Accelerometers. In Proceedings of the Fourth International Workshop on Knowledge Discovery from Sensor Data (SensorKDD-10) at KDD-10, Washington, DC, USA.

• Malekzadeh, M., Clegg, R. G., Cavallaro, A., & Haddadi, H. (2019). Mobile Sensor Data Anonymization. In Proceedings of the International Conference on Internet of Things Design and Implementation (IoTDI '19), Montreal, Quebec, Canada (pp. 49–58). ACM. https://doi.org/10.1145/3302505.3310068

• Reyes-Ortiz, J. L., Anguita, D., Ghio, A., Oneto, L., & Parra, X. (2013). Human Activity Recognition Using Smartphones. UCI Machine Learning Repository. https://doi.org/10.24432/C54S4K

---

✨ Author

Developed by Stefanos Konstantinou for Challenge 2025 – ADAMMA.