PlateVision Nepal 🇳🇵

AI-Powered Automatic Number Plate Recognition (ANPR / ALPR) for Nepal

A production-grade, end-to-end Vehicle License Plate Detection and Recognition system specifically engineered for Nepal's transport ecosystem. It accurately detects vehicles (cars, motorcycles, scooters, buses, trucks) and recognizes license plates in both Nepali (Devanagari script) and English (Latin High-Security Embossed script).

---

🌟 Key Capabilities

1. Dual-Script Recognition:

   • Devanagari Script (नेपाली):
     • Traditional Zonal format: e.g. बा ९२ प ५६७८ (Bagmati Two-Wheeler), ना ५ ख १२३४ (Narayani Public Bus), बा १५ च ४३२१ (Private Car), बा १ ज ७८९० (Taxi)
     • Provincial format: e.g. बा.प्र. ०२-०३५ प ५६७८ (Bagmati Province), प्रदेश ३ - ०१ ०२५ च ९८७६
     • Stacked 2-line layout parsing (common on motorcycles, tempos, and trucks)
   • English Script (High-Security Embossed):
     • Provincial Embossed format: e.g. BAGMATI PROVINCE 02-035 PA 5678, 02-035 PA 5678, BA 2 PA 1234
     • Classification mapping: PA/PB/PC (Two-wheelers), CA/CB (Light 4-wheelers), BA/BB (Heavy vehicles), DA (Three-wheelers)

2. Multi-Vehicle Classification:

   • YOLOv8 deep learning model recognizing car, motorcycle, bus, and truck with bounding box annotations.

3. Multi-Tier Plate Localization Pipeline:

   • Tier 1 (Custom YOLOv8): Uses models/license_plate_detector.pt if present.
   • Tier 2 (OpenCV Haar Cascade): Multi-scale plate classifier fallback.
   • Tier 3 (Nepal Color Segmenter): Segments by Nepal's official color schemes:
     • 🔴 Red plate: Private vehicles (निजी सवारी)
     • ⚫ Black plate: Public transport, buses, trucks, taxis (सार्वजनिक सवारी)
     • ⚪ White plate: High-security embossed and government vehicles (इम्बोस्ड / सरकारी)
     • 🟢 Green plate: Electric vehicles (विद्युतीय सवारी EV)
   • Tier 4 (Morphological & Gradient): Vertical Sobel edge filter + rectangular aspect ratio filtering.

4. Image Enhancement & Preprocessing:

   • CLAHE (Contrast Limited Adaptive Histogram Equalization) on LAB color space.
   • Bilateral edge-preserving denoising.
   • Bicubic super-resolution scaling for low-resolution or distant crops.

5. Modern Glassmorphic UI Dashboard:

   • Frosted glass cards with backdrop-filter: blur(24px).
   • Dark luxury theme with neon cyan, emerald green, and amber highlights.
   • Interactive Tabs:
     • 🔍 Live Scanner & Upload: Drag & drop or 1-click Quick Demo testing.
     • 📹 Live Stream / Webcam: Real-time video processing with live plate feed.
     • 📋 Plate Registry: Searchable database, crop image preview lightbox, CSV export.
     • 📊 Analytics: Interactive Chart.js graphs (vehicle breakdown, script ratio, hourly flow).
     • 🇳🇵 Nepal Plate Guide: Comprehensive educational guide on Nepal number plate formats.

---

📁 Project Architecture

plateVision/
├── main.py                   # FastAPI backend server & REST API endpoints
├── detector.py               # Production ANPR engine (YOLOv8 + EasyOCR + CV Segmenter)
├── nepal_plate_parser.py     # Domain-specific Nepal plate parsing, transliteration & rules
├── database.py               # SQLite database management with duplicate suppression & stats
├── generate_samples.py       # Realistic test sample generator (Bikes, Cars, Buses, Embossed)
├── requirements.txt          # Python dependencies
├── static/
│   ├── css/
│   │   └── style.css         # Glassmorphism styling, responsive layout & neon accents
│   ├── js/
│   │   └── app.js           # Frontend client engine (Uploads, Webcam, Charts, Modals)
│   ├── crops/                # Stored plate crops for instant lightbox inspection
│   └── samples/              # One-click demo sample images
├── templates/
│   └── index.html            # HTML5 Dashboard with modern glassmorphism design
└── data/
    └── anpr.db               # SQLite persistent database

---

🚀 Running the Application

1. Activate the Virtual Environment

source venv/bin/activate

2. Start the FastAPI Web Server

python -m uvicorn main:app --host 127.0.0.1 --port 8001

3. Open in Browser

Visit:

http://127.0.0.1:8001

---

📡 REST API Endpoints

• GET / : Interactive Glassy Web Dashboard
• POST /api/detect/image : Upload image (multipart/form-data) for ANPR
• POST /api/detect/sample : Analyze preloaded sample (sample_name=sample_bike_nepali.jpg)
• POST /api/detect/frame : Base64 video frame processing for webcam / RTSP CCTV
• GET /api/history : Get paginated and filtered plate records (?search=...&vehicle_type=...&script=...)
• GET /api/stats : Real-time analytics and fleet counts
• GET /api/export/csv : Download all historical detections in CSV format
• DELETE /api/detections/{id} : Delete specific record
• POST /api/detections/clear : Clear all detection records
• GET /api/health : System diagnostics, compute device (MPS/CUDA/CPU), and model states