Open-source vehicle detection system using Edge AI, ESP32-CAM, and FOMO (Faster Objects, More Objects)
Monitor traffic flow in real-time using affordable IoT hardware and machine learning at the edge.
Real-time traffic monitoring dashboard featuring:
- Full Corridor Visualisation - CBD to Fremantle (3 stretches, 6 km)
- Speed Estimation - Real-time traffic flow analysis using density theory
- Interactive Map - Colour-coded routes showing congestion levels (Green → Orange → Red → Dark Red)
- Bidirectional Tracking - Separate Northbound/Southbound monitoring
- Hourly Charts - Historical patterns across 24 monitoring sites
- WA-Themed UI - Cottesloe Beach and Indigenous Earth colour schemes (light/dark modes)
Main Roads Western Australia operates sophisticated Smart Freeway infrastructure with 1,400+ sensors — a significant investment that requires coordination across departments, professional installation, ongoing maintenance, and integration with broader transport systems.
SwanFlow explores how citizen-led monitoring might complement official infrastructure by providing additional data points on arterial roads that may fall outside freeway sensor coverage.
This project is a technical proof-of-concept and a gentle advocacy tool for greater open data access.
We believe that combining official infrastructure with citizen contributions could create richer traffic datasets for researchers, urban planners, and commuters alike.
📖 See docs/mainroads-api-investigation.md for our research into public API availability.
SwanFlow is a DIY traffic monitoring system that:
- Monitors 6 km arterial corridor from CBD to Fremantle (3 stretches)
- 18 monitoring sites with bidirectional tracking (Northbound/Southbound)
- Real-time speed estimation using traffic flow theory
- Edge AI detection using Edge Impulse FOMO (ML model on ESP32)
- Closed-segment monitoring for accurate flow measurement
- Costs ~$223 AUD per monitoring site (solar-powered)
- Runs 24/7 completely off-grid with solar power
Perfect for:
- Commuter intelligence: Real-time arterial road monitoring
- Open data advocacy: Demonstrating value of accessible traffic data
- Proof of concept: Closed-segment monitoring validation
- Research projects (urban planning, transport optimisation)
- Hobbyists (IoT, ML, embedded systems)
- Community projects (local traffic advocacy)
- ESP32-CAM (OV2640 camera) - $8-12
- SIM7000A (4G LTE module) - $15-20
- 20W Solar Panel + 12V Battery - ~$80
- Edge Impulse FOMO (on-device ML inference)
- MicroSD for local image buffering
- Weatherproof (IP65 junction box)
- 100% off-grid with solar power (no mains required)
- PlatformIO (Arduino framework)
- Real-time vehicle counting
- Configurable detection thresholds
- SD card logging
- LTE data upload (via TinyGSM)
- Watchdog and error recovery
- Express.js REST API
- SQLite database
- Aggregated statistics (hourly, daily, monthly)
- Authentication (API key)
- CORS-enabled for dashboard
- Real-time dashboard (Chart.js)
- Traffic flow visualisation
- Multi-site support
- Mobile-responsive
- Auto-refresh (60s interval)
- Knowledge Base — Technical documentation with interactive cards and Lucide icons
See hardware/shopping-lists.md for complete shopping list.
AliExpress (2-4 week shipping):
- 2x ESP32-CAM (OV2640)
- 1x SIM7000A LTE module
- 1x USB-to-TTL programmer
- 2x MicroSD cards (8-16GB)
- 1x 20W solar panel
- 1x 12V 7Ah battery
- 1x Solar charge controller
Bunnings (same day):
- IP65 junction box
- Cable glands
- Mounting bracket
- Solar panel mounting hardware
M2M SIM Card:
- Register at m2msim.com.au
- 1GB/month plan (~$5-8/month)
Total Budget: ~$223 AUD (Phase 1 - Solar-Powered)
- Create account at Edge Impulse Studio
- Collect 300-500 training images (use phone or ESP32-CAM)
- Label vehicles in Edge Impulse
- Train FOMO model (see docs/ml-development-guide.md)
- Export as Arduino library
cd firmware/esp32-cam-counter
# Install PlatformIO
pip install platformio
# Update config.h with your settings
# - WiFi credentials (for initial testing)
# - M2M APN settings
# - Server URL
# - API key
# Build and upload
pio run -t upload
# Monitor serial output
pio device monitorOption A: Render.com (Recommended - FREE)
The easiest and cheapest option - completely free hosting with auto-deployment:
- Create account at render.com (free tier)
- Connect your GitHub repository
- Create new "Web Service" from
render.yaml - Set environment variable:
API_KEY=your_secret_key - Deploy automatically on every GitHub push
Free Tier Includes:
- 750 hours/month (24/7 coverage)
- HTTPS included
- Auto-deploy from GitHub
- 0.1 CPU, 512MB RAM (perfect for this project)
- Total cost: $0/month
Backend will be live at: https://your-app.onrender.com
Option B: Local Development
cd backend/api
# Install dependencies
npm install
# Configure environment
cp .env.example .env
nano .env # Set API_KEY
# Run server
npm startBackend runs on http://localhost:3000
Option A: Vercel (Recommended - FREE)
Deploy dashboard for free with HTTPS:
cd frontend/web-dashboard
# Install Vercel CLI
npm install -g vercel
# Deploy (first time)
vercel
# Deploy updates
vercel --prodFree Tier Includes:
- Unlimited bandwidth
- HTTPS/SSL included
- 100GB bandwidth
- Auto-deploy from GitHub
- Total cost: $0/month
Dashboard will be live at: https://your-app.vercel.app
Option B: Local Development
cd frontend/web-dashboard
# Serve locally (Python)
python -m http.server 8080Dashboard accessible at http://localhost:8080
- Complete docs/site-survey-checklist.md
- Install junction box and mount ESP32-CAM
- Power on and verify LTE connection
- Monitor dashboard for 24 hours
Live Site: swanflow.com.au (Vercel) → Backend: perth-traffic-watch.onrender.com (Render.com)
The live dashboard currently runs on simulated traffic data while hardware is being prepared:
Backend (Render.com Free Tier):
- Express.js API generating realistic traffic patterns for all 26 monitoring sites
- Simulates bidirectional flow (Northbound/Southbound) with time-of-day variations
- SQLite database stores simulated counts and statistics
- Cost: $0/month (within 750 hours/month free tier limit)
Keep-Alive Mechanism (Zero Cost):
- Primary: Cron job on Donnacha VPS (45.77.233.102) pings backend every 12 minutes
*/12 * * * * curl -s -X HEAD https://perth-traffic-watch.onrender.com/api/sites
- Redundancy: Frontend JavaScript pings backend when users have dashboard open
- Result: Backend stays warm 24/7, users get instant data loads (1-2 seconds instead of 30+ second cold starts)
Why Keep-Alive?
- Render.com free tier "sleeps" after 15 minutes of inactivity
- Without keep-alive: 30+ second wait for first load
- With keep-alive: Instant data every time
Simulation → Real Data Transition:
- Backend API endpoints remain identical
- When ESP32-CAM devices are deployed, they'll POST real counts to same endpoints
- Dashboard seamlessly switches from simulated to real data
- No frontend changes required
Cost Summary:
- Backend hosting: $0 (Render.com free tier)
- Keep-alive cron job: $0 (runs on existing Donnacha VPS)
- Frontend hosting: $0 (Vercel free tier)
- Total: $0/month 🎉
swanflow/
├── hardware/
│ ├── bom.md # Bill of materials
│ └── shopping-lists.md # AliExpress + Bunnings orders
│
├── firmware/
│ └── esp32-cam-counter/
│ ├── platformio.ini # PlatformIO config
│ └── src/
│ ├── main.cpp # Main loop
│ ├── config.h # All configuration
│ ├── vehicle_counter.* # FOMO detection + counting
│ └── lte_modem.* # SIM7000A communication
│
├── backend/
│ └── api/
│ ├── package.json
│ ├── index.js # Express API + SQLite
│ └── README.md # Backend setup guide
│
├── frontend/
│ └── web-dashboard/
│ ├── index.html
│ ├── styles.css
│ ├── app.js # Dashboard logic (Chart.js)
│ ├── knowledge.html # Knowledge Base page
│ ├── knowledge.css # Knowledge page styles
│ ├── knowledge.js # Knowledge page interactions
│ └── README.md # Frontend setup guide
│
├── tests/
│ └── e2e/
│ ├── playwright.config.js # Playwright configuration
│ ├── pages/ # Page Object Models
│ │ ├── DashboardPage.js
│ │ └── KnowledgePage.js
│ ├── dashboard.spec.js # Dashboard tests
│ ├── knowledge.spec.js # Knowledge page tests
│ ├── mobile-viewport.spec.js
│ ├── accessibility.spec.js
│ ├── performance.spec.js
│ ├── visual-regression.spec.js
│ └── README.md # Test documentation
│
├── docs/
│ ├── corridor-architecture.md # 3-stretch system architecture
│ ├── cost-effectiveness-analysis.md # Hardware cost breakdown
│ ├── mainroads-api-investigation.md # API outage research
│ ├── academic-paper-plan.md # Research publication roadmap
│ ├── ml-development-guide.md # Edge Impulse training guide
│ ├── iot-sim-options.md # M2M SIM comparison (Australia)
│ ├── requirements-and-todos.md # Project roadmap
│ ├── legal-considerations.md # Privacy, permits, compliance
│ ├── site-survey-checklist.md # Field survey template
│ └── contributing.md # How to contribute
│
└── README.md # This file
Primary Approach: Bidirectional Co-Located (RECOMMENDED)
- 2 locations with co-located cameras = 4 sensors
- Hardware per location: 2 × ESP32-CAM modules sharing solar panel, battery, and mounting infrastructure
- Cost: $540 total ($270 per location)
- Example: Crawley + Point Lewis, each with 2 cameras (inbound + outbound)
- Why This Approach:
- ✅ Tests complete bidirectional flow measurement
- ✅ 40% cheaper than separate installations ($892)
- ✅ Only $94 more than unidirectional ($446)
- ✅ Easier permitting (can approach private property owners)
- ✅ Modular design: each camera fully independent, shares power only
- ✅ Single mounting structure per location reduces installation complexity
Hardware Breakdown per Co-Located Location ($270):
- Shared infrastructure ($150): 20W solar panel ($35), 12V battery ($30), charge controller ($15), large junction box ($50), mounting bracket ($20)
- Per-camera modules (×2 = $120): ESP32-CAM ($12), SIM7000A 4G modem ($20), MicroSD card ($8), programmer ($10), wiring/cables ($10)
Alternative: Unidirectional Prototype (Documentation Only)
- 2 locations × 1 direction = 2 installations at $223 each = $446
- Monitors one direction only (e.g., outbound to Fremantle)
- Simpler but doesn't test full bidirectional capability
Crawley → Point Lewis (~1.5 km)
- Status: Proof of Concept Target
- Monitoring Sites: Crawley (entry), Point Lewis (exit) — 2 co-located installations
- Characteristics: Waterfront arterial, zero side access, perfect closed segment
- Why Optimized: True closed segment needs only entry/exit points
- PoC Cost: $540 (2 co-located locations, 4 cameras total, both directions)
Grant St → Eric St (~1.5 km)
- Status: Future Expansion
- Monitoring Sites: Grant St, Campbell Barracks, Eric St (3 × 2 directions = 6 sensors)
- Characteristics: Campbell Barracks creates natural barrier, very few side access points
- Key Feature: Army facility on one side = minimal civilian traffic entry/exit
Forrest St → Victoria St (~3 km)
- Status: Future Expansion
- Monitoring Sites: Forrest St, Bay View Terrace, McCabe St, Victoria St (4 × 2 directions = 8 sensors)
- Characteristics: Longest stretch, residential arterial, tests algorithm with side streets
- Purpose: Validate algorithm with more complex access patterns
Total: 18 monitoring sites (9 locations × 2 directions each) across 3 arterial stretches
See: docs/corridor-architecture.md for detailed technical architecture
Main Roads WA operates extensive traffic monitoring infrastructure. Our research found:
| Observation | Detail |
|---|---|
| Infrastructure | 1,400+ Smart Freeway sensors collecting real-time data |
| Internal Systems | Sophisticated RTOP platform for traffic management |
| Public APIs | Currently unavailable (since August 2023) |
| Operations | 24/7 Operations Centre managing traffic flow |
We understand that maintaining public-facing APIs requires resources, security considerations, and ongoing support — challenges that may not be immediately visible to outside observers.
We'd love to see more open traffic data available to researchers and the public:
- Public API restoration — when resources and security concerns allow
- Arterial road coverage — where freeway sensors may not reach
- Citizen-augmented data — complementing (not replacing) official sources
- Following leaders like NSW — Transport for NSW provides excellent open data as a model
See: docs/mainroads-api-investigation.md for our research.
Camera → FOMO Inference → Counting Algorithm → Stats
- ESP32-CAM captures frame (QVGA 320x240)
- Edge Impulse FOMO detects vehicles (on-device)
- Counting algorithm tracks vehicles crossing virtual line
- Accumulates stats: total count, hourly count, confidence
ESP32 → SIM7000A → 4G LTE → Backend API
- Every 60 seconds, upload JSON stats to backend
- Optional: Upload detection images for validation
- Uses ~200-500MB/month data
API → SQLite → Aggregation → Dashboard
- Stores detection events in SQLite
- Aggregates statistics (hourly, daily)
- Serves data to web dashboard via REST API
Vehicle Counts → Traffic Flow Theory → Estimated Speed
(Per stretch) (Flow ÷ Density) (5-65 km/h)
- Calculate average vehicle flow per stretch per direction
- Apply traffic flow theory: Speed = Flow ÷ Density
- Calibrated for 60 km/h arterial roads
- Colour-code routes: Green (flowing) → Orange (moderate) → Red (heavy) → Dark Red (gridlock)
See: docs/corridor-architecture.md for algorithm details and evolution roadmap
Dashboard → Fetch API → Chart.js → User
- Real-time traffic statistics across 3 stretches
- Colour-coded route visualisation (6 routes: 3 stretches × 2 directions)
- Hourly traffic flow chart with speed estimates
- "Should I drive now?" intelligence
- Auto-refresh every 60 seconds
Target: >70% detection accuracy in real-world conditions
Factors affecting accuracy:
- Good lighting (daytime, well-lit roads)
- Poor lighting (night, heavy shadows)
- Clear camera view
- Occlusions (trees, rain, fog)
- Medium traffic density
- Very high density (overlapping vehicles)
Validation: Manual count vs. automated count over 1 hour
The figures below represent hardware-only costs for a hobbyist/citizen project. They don't include:
- Professional labour and installation
- Permitting and compliance
- Long-term maintenance contracts
- Integration with existing systems
- Metal fabrication and weatherproofing to commercial standards
- Interdepartmental coordination
Government infrastructure serves different purposes (reliability, coverage, legal compliance, integration) and direct cost comparisons aren't meaningful. SwanFlow is a complementary proof-of-concept, not a replacement for professional traffic monitoring.
| Component | Cost (AUD) |
|---|---|
| ESP32-CAM | $12 |
| SIM7000A LTE module | $20 |
| 20W Solar panel | $35 |
| 12V 7Ah battery | $30 |
| Solar charge controller | $15 |
| MicroSD card | $8 |
| USB-to-TTL programmer | $10 |
| Junction box + mounting | $40 |
| Solar mounting hardware | $15 |
| Cables, glands, misc | $30 |
| M2M SIM (first month) | $8 |
| Total | ~$223 |
Key Advantages:
- ✅ 100% off-grid - No mains power required
- ✅ Zero power costs - Free energy from the sun
- ✅ Deploy anywhere - Not limited by power outlet locations
- ✅ Storm resilient - Continues operating during power outages
- Bulk discounts: ~$200/site
- Data pooling: Share 5GB across 5 devices (~$12/month total)
- Total: ~$1,000 hardware + $12/month
Backend/Frontend Hosting: $0 (Render.com + Vercel free tiers)
Advantages at Scale:
- ✅ No trenching costs - No power cable installation
- ✅ No electrician fees - Self-contained solar system
- ✅ Flexible placement - Deploy anywhere with sunlight
- ✅ Easy relocation - Move sites without rewiring
| Deployment | Hardware | Monthly Cost | Notes |
|---|---|---|---|
| Phase 0: Bidirectional PoC (Co-Located) | $540 | $8 | 2 co-located locations, 4 sensors, both directions (RECOMMENDED START) |
| Phase 1: + Swanbourne (Co-Located) | $1,350 additional ($1,890 total) | $12 | 3 co-located locations added (6 sensors) |
| Phase 2: Full Corridor | $2,124 additional ($4,014 total) | $15 | 18 sensors, complete arterial monitoring |
| Backend + Frontend Hosting | $0 | $0 | Render.com + Vercel (FREE) |
Key Cost Advantages:
- ✅ No hosting costs - Render.com & Vercel free tiers
- ✅ No server costs - Fully cloud-hosted
- ✅ No power costs - 100% solar-powered
- ✅ No electrician costs - Self-contained solar system
- ✅ No SSL cert costs - HTTPS included
- ✅ Auto-deployment - Push to GitHub = instant deploy
Compare to Traditional Setup:
- VPS: ~$5-15/month
- SSL Certificate: ~$50-200/year
- Power bill: ~$5-10/month per site
- Electrician for install: ~$200-500 per site
- SwanFlow: $0/month for hosting & power 🎉
Cost: $540 | Timeline: 2-4 weeks
- Hardware BOM and shopping list
- Firmware structure (PlatformIO)
- Backend API (Express + SQLite)
- Web dashboard (Chart.js)
- Documentation
- Edge Impulse model training
- Order hardware for 2 co-located locations (Crawley + Point Lewis)
- Configure firmware for bidirectional monitoring
- Build co-located mounting infrastructure (shared solar/battery)
- Field deployment: 2 locations, 4 cameras (2 per location)
- Validate bidirectional flow measurement
Success Criteria:
- ✅ Inbound vehicles at Crawley ≈ Outbound vehicles at Point Lewis (±10%)
- ✅ Outbound vehicles at Crawley ≈ Inbound vehicles at Point Lewis (±10%)
- ✅ Cost under $600
- ✅ System runs 24/7 on solar power
- ✅ Co-located cameras operate independently
Cost: $1,350 additional ($1,890 total)
- Survey Swanbourne stretch (Grant St → Eric St)
- Install 3 additional co-located locations (6 sensors)
- Data pooling (shared SIM plan)
- Validate algorithm with multiple stretches
Cost: $2,124 additional ($4,014 total)
- Complete Mosman Park stretch (4 locations, 8 sensors)
- 18 total sensors across 6 km
- Complete arterial corridor monitoring
- Multi-class detection (cars, trucks, motorcycles)
- Improved speed estimation algorithms
- Alerts (email/SMS for congestion)
- Historical data analysis
- Mobile-responsive dashboard improvements
- API documentation for third-party access
See docs/requirements-and-todos.md for full roadmap.
| Document | Description |
|---|---|
| docs/corridor-architecture.md | 3-stretch system architecture, algorithm details, ML integration |
| docs/academic-paper-plan.md | Research publication roadmap: structure, venues, contributions |
| docs/mainroads-api-investigation.md | Research into public traffic data availability in WA |
| hardware/bom.md | Bill of materials |
| hardware/shopping-lists.md | AliExpress + Bunnings orders |
| docs/ml-development-guide.md | Edge Impulse FOMO training |
| docs/iot-sim-options.md | M2M SIM comparison (Australia) |
| docs/requirements-and-todos.md | Project roadmap and TODOs |
| docs/legal-considerations.md | Privacy, permits, compliance |
| docs/site-survey-checklist.md | Site evaluation template |
| docs/contributing.md | How to contribute |
| backend/README.md | Backend API setup |
| frontend/README.md | Dashboard setup |
| tests/e2e/README.md | E2E test suite: 120+ tests, Playwright, accessibility, performance |
| frontend/web-dashboard/knowledge.html | Knowledge Base: Interactive documentation with Lucide icons |
| CHANGELOG.md | Version history and release notes |
A: Recording vehicles on public roads is generally legal in Australia (WA Surveillance Devices Act 2016). However:
- Don't record audio (requires consent)
- Don't record private property
- May need permit from City of Perth for installation
See docs/legal-considerations.md for details.
A: Target is >70% accuracy. Actual accuracy depends on:
- Camera angle and positioning
- Lighting conditions
- Traffic density
- Model training quality
Validate with manual counts.
A: SwanFlow:
- Uses low-resolution images (QVGA 320x240)
- No facial recognition
- No number plate recognition
- No audio recording
- Anonymous vehicle counts only
A: Yes! MIT License allows commercial use. However:
- Consult legal advice for privacy compliance
- May need public liability insurance
- Check local regulations
A: Estimated ~200-500MB/month per device:
- Stats only: ~9MB/month
- Stats + occasional images: ~150MB/month
A: Depends on street lighting:
- Well-lit roads: Yes (requires good ambient light)
- Poorly lit roads: May need IR illumination
- Test at night before deployment
A: Yes! ESP32-CAM has built-in WiFi:
- Cheaper (no SIM card)
- Requires WiFi access point nearby
- Update firmware to use WiFi instead of LTE
A: Solar makes SwanFlow truly independent:
- No power outlet needed - Deploy anywhere
- Zero ongoing power costs - Free energy
- Storm resilient - Continues during outages
- Easy relocation - No rewiring required
- Lower total cost - No electrician fees (~$200-500/site saved)
SwanFlow includes a comprehensive E2E test suite built with Playwright, covering the live production dashboard at swanflow.com.au.
# Install dependencies
npm install
# Run all tests
npm test
# Run with UI
npm run test:ui| Command | Description |
|---|---|
npm test |
Run all 120+ tests |
npm run test:chrome |
Desktop Chrome only |
npm run test:mobile |
iPhone + Android viewports |
npm run test:a11y |
Accessibility (WCAG 2.1 AA) |
npm run test:perf |
Performance & Web Vitals |
npm run test:visual |
Visual regression screenshots |
npm run test:dashboard |
Dashboard functionality |
npm run test:knowledge |
Knowledge page |
- Dashboard Tests — Theme switching, network tabs, map controls, terminal, journey visualization
- Knowledge Page — Card interactions, filtering, navigation
- Mobile Viewports — iPhone SE/13/14 Pro Max, Pixel 5, Galaxy S9+, iPad
- Accessibility — WCAG 2.1 AA compliance, keyboard navigation, color contrast
- Performance — Core Web Vitals (LCP, FID, CLS), load times, memory usage
- Visual Regression — Full-page and component screenshots for both themes
- Network/API — Request validation, mocking, error handling
- Links — All internal/external links validated
| Metric | Target | Description |
|---|---|---|
| LCP | < 2.5s | Largest Contentful Paint |
| FID | < 100ms | First Input Delay |
| CLS | < 0.1 | Cumulative Layout Shift |
| FCP | < 1.8s | First Contentful Paint |
| TTFB | < 800ms | Time to First Byte |
See tests/e2e/README.md for full documentation.
| Component | Technology |
|---|---|
| Hardware | ESP32-CAM (OV2640), SIM7000A, MicroSD |
| Firmware | Arduino (ESP32), PlatformIO, TinyGSM |
| ML Model | Edge Impulse FOMO, TensorFlow Lite Micro |
| Backend | Node.js, Express.js, SQLite, better-sqlite3 |
| Frontend | Vanilla JS, Chart.js, HTML5, CSS3 |
| Testing | Playwright, axe-core (accessibility) |
| Deployment | Vercel (frontend - FREE), Render.com (backend - FREE) |
Contributions welcome! See docs/contributing.md for guidelines.
Ways to contribute:
- Report bugs
- Suggest features
- Improve documentation
- Field testing
- ML model improvements
- Hardware alternatives
MIT License - see LICENSE for details.
Free to use, modify, and distribute. No warranty.
- Edge Impulse - FOMO object detection
- Espressif - ESP32 platform
- TinyGSM - Modem library
- Chart.js - Dashboard visualisation
- GitHub Issues: Report bugs and request features
- GitHub Discussions: Ask questions and share ideas
- Edge Impulse: https://edgeimpulse.com
- ESP32-CAM Community: https://github.com/espressif/esp32-camera
Built with love in Perth, Western Australia
Open-source traffic monitoring for hobbyists and researchers. Gentle advocacy for more open data.
- ⭐ Star this repository
- 📖 Read docs/mainroads-api-investigation.md to understand the open data landscape
- 🛒 Order hardware (see hardware/shopping-lists.md)
- 📸 Collect training images
- 🤖 Train FOMO model (see docs/ml-development-guide.md)
- 🚀 Deploy and share your results!
Questions? Open an issue or discussion on GitHub.
This project is being developed with academic publication in mind. Key documents:
- docs/academic-paper-plan.md — Full research roadmap
- docs/mainroads-api-investigation.md — Research into public data availability
Research Contributions:
- Closed-Segment Monitoring Theory — Entry/exit point tracking for arterial roads
- Citizen-Augmented Monitoring Framework — How hobbyist projects can complement official infrastructure
- Open Data Accessibility Study — Comparing approaches across Australian states
Status: Proof of Concept (Phase 1) Version: 0.2.1 Last Updated: 2025-12-26