Support my work

If this project has helped you, consider supporting my work on Buy Me a Coffee.

RadiaCode Data Collection & Isotope Identification

A complete system for radiation detection, data collection, and AI-powered isotope identification using RadiaCode scintillation detectors.

Projects

This monorepo contains three interconnected applications:

1. Open RadiaCode Android App (android_app/)

A Kotlin Android app that connects to RadiaCode devices over Bluetooth LE.

Features:

• Real-time dose rate and count rate monitoring
• Live charts with zoom, pan, and statistical analysis
• Home screen widgets with customizable themes
• Smart alerts with threshold and statistical triggers
• Live GPS mapping of radiation readings
• Real-time isotope identification via backend API
• Background service with auto-connect on boot

Full documentation

2. Vega Middleware (middleware/)

A suite of AI microservices deployed on GPU servers via Docker.

Service	Port	Description
vega-tts	8000	Text-to-Speech with voice cloning (Chatterbox TTS)
vega-llm	8001	Chat and text generation (Qwen)
vega-isotope-identification	8020	CNN-based isotope identification
vega-ingress	8080	API gateway with request logging

Full documentation

3. Vega ML Training (vega_ml/)

Machine learning pipeline for training isotope identification models.

Components:

• Synthetic gamma spectra generator (82 isotopes)
• Physics-based simulation (Gaussian peaks, Poisson noise, Compton continuum)
• CNN-FCNN hybrid model (VegaModel, 34.5M parameters)
• Training pipeline with mixed precision and multi-task learning

Full documentation

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                         User's Phone                            │
│  ┌─────────────────────────────────────────────────────────┐   │
│  │              Open RadiaCode Android App                  │   │
│  │  - Dashboard with live charts                           │   │
│  │  - Widgets, alerts, GPS mapping                         │   │
│  │  - Isotope identification UI                            │   │
│  └─────────────────────────────────────────────────────────┘   │
└──────────────────────────┬──────────────────────────────────────┘
                           │ BLE                    │ HTTPS
                           ▼                        ▼
                    ┌─────────────┐         ┌──────────────────┐
                    │  RadiaCode  │         │  Vega Middleware │
                    │   Device    │         │  (GPU Server)    │
                    │  (101-110)  │         │                  │
                    └─────────────┘         │  ┌────────────┐  │
                                            │  │ Isotope ID │  │
                                            │  │   (CNN)    │  │
                                            │  └────────────┘  │
                                            │  ┌────────────┐  │
                                            │  │    LLM     │  │
                                            │  └────────────┘  │
                                            │  ┌────────────┐  │
                                            │  │    TTS     │  │
                                            │  └────────────┘  │
                                            └──────────────────┘
                                                     ▲
                                                     │ model files
                                            ┌────────┴─────────┐
                                            │    Vega ML       │
                                            │  Training        │
                                            │  (RTX 5090)      │
                                            └──────────────────┘

Quick Start

Android App

1. Download the latest APK from android_app/Installer/
2. Enable "Install from unknown sources" on your Android device
3. Install and grant Bluetooth/Location permissions
4. Pair with your RadiaCode device

Middleware (Server Deployment)

# Deploy isotope identification service
cd middleware/vega-isotope-identification
cp .env.example .env  # Edit with your server details
.\deploy.ps1

# Deploy TTS service
cd ../vega-tts
cp .env.example .env
.\deploy.ps1

ML Training

cd vega_ml

# Create virtual environment
python -m venv .venv
.venv\Scripts\activate  # Windows

# Install dependencies
pip install numpy scipy pillow
pip install --pre torch --index-url https://download.pytorch.org/whl/nightly/cu128

# Generate training data
python -m synthetic_spectra.generate_spectra

# Train model
python training/vega/run_training.py --epochs 100

Supported Hardware

RadiaCode Devices

Model	Crystal	FWHM @ 662 keV	Status
RadiaCode 101	CsI(Tl)	9.0%	Supported
RadiaCode 102	CsI(Tl)	9.5%	Supported
RadiaCode 103	CsI(Tl)	8.4%	Supported
RadiaCode 103G	GAGG(Ce)	7.4%	Supported
RadiaCode 110	CsI(Tl)	8.4%	Primary

GPU Requirements

• Training: NVIDIA RTX 3090/4090/5090 (24GB VRAM recommended)
• Inference: RTX 3090 or equivalent for real-time isotope ID

Isotopes Supported

The system can identify 82 isotopes across categories:

• Natural Background: K-40, Ra-226, Rn-222, Th-232, U-238
• Medical: Tc-99m, F-18, I-131, Ga-68, Lu-177
• Industrial: Ir-192, Se-75, Am-241, Cs-137, Co-60
• Calibration: Ba-133, Eu-152, Na-22, Mn-54, Co-57
• Reactor Fallout: Cs-134, Sr-90, Zr-95, Ru-106

Development

Prerequisites

• Android Studio (for Android app)
• Python 3.10+ (for ML and middleware)
• Docker with NVIDIA Container Toolkit (for deployment)
• ADB (for device testing)

Project Structure

RadiaCodeAndroidDataCollection/
├── android_app/          # Kotlin Android application
│   ├── app/src/          # Source code
│   ├── Installer/        # Distributable APKs
│   └── AGENTS.md         # Android-specific agent instructions
│
├── middleware/           # AI microservices
│   ├── vega-tts/         # Text-to-Speech service
│   ├── vega-llm/         # LLM chat service
│   ├── vega-isotope-identification/  # CNN inference service
│   ├── vega-ingress/     # API gateway
│   └── agent.md          # Middleware agent instructions
│
├── vega_ml/              # ML training pipeline
│   ├── synthetic_spectra/  # Data generation
│   ├── training/         # Model and training code
│   ├── inference/        # Inference engine
│   ├── models/           # Saved checkpoints
│   └── agents.md         # ML agent instructions
│
├── AGENTS.md             # Root agent instructions
└── README.md             # This file

Agent Instructions

For AI coding agents, see AGENTS.md for operating rules and per-project documentation.

License

See individual project folders for license information.