DeepVision — Face + Voice Authentication Suite

DeepVision is a finished, end‑to‑end biometric authentication project that combines face recognition, voice verification, and real‑time attention monitoring into a single interactive system. The goal is to provide a practical, demo‑ready pipeline that can be run locally or backed by a cloud database.

Highlights

• Multi‑modal authentication: face recognition + voice verification for stronger identity checks.
• Real‑time monitoring: eye aspect ratio (EAR) and head‑pitch cues for attention/drowsiness signals.
• Gradio UI: a ready‑to‑run interface for demos and validation.
• Two storage backends:
  • Local SQLite for offline/demo usage.
  • Azure SQL for cloud deployments.

Project structure

.
├── mygradio2.py           # Main application (SQLite-backed)
├── mygradio.py            # Azure SQL variant (pyodbc)
├── Models/                # Voice verification models (.h5)
├── persons/               # Reference face images (name-based)
├── voices/                # Reference voice samples (per-person folders)
├── Database/              # SQLite database (faces.db)
├── test.py                # Voice model evaluation/threshold script
└── requirements*.txt      # Python dependencies

How the system works (at a glance)

1. Face detection + embedding
   MTCNN detects faces and FaceNet generates embeddings.
2. Face match
   Embeddings are compared against the database via cosine similarity.
3. Voice verification
   A Siamese model produces voice embeddings for similarity checks.
4. Attention monitoring
   Eye aspect ratio and head pitch provide basic drowsiness cues.

Setup (local, finished workflow)

1. Install dependencies

   pip install -r requirements.txt

2. Add reference data

   • Put face images in persons/ named like person_name.jpg.
   • Put voice samples in voices/person_name/*.wav.

3. Run the main app

   python mygradio2.py

This launches the full Gradio UI and automatically seeds the local database (Database/faces.db).

Azure SQL deployment

Use mygradio.py when you want a cloud-backed deployment:

python mygradio.py

Note: You must provide valid Azure SQL credentials and have ODBC drivers installed.

Model notes

• The system expects a trained voice model at:

  Models/Voice_verification_model5.h5
  

• Additional model versions are stored in Models/ for experimentation.

Evaluation & validation

Use test.py to evaluate the voice model, compute a similarity threshold, and export misclassifications:

python test.py

Status

✅ Complete — This repository represents a finished, working project with a runnable UI, local/cloud database support, and evaluation utilities.

---

If you are new to the code, start with mygradio2.py to see the full, working pipeline from capture → recognition → verification → monitoring.