Real-time vehicle intelligence

Shahin

High-fidelity vehicle + plate tracking for safety-critical, edge-ready deployments.

99.1% Plate recognition uptime

+200 Vehicles tracked per second

Edge & cloud Deployment options

Precision you can trust

About the Project

This modular orchestration captures vehicles and their license plates with laser-focused accuracy, funneling live video into state-of-the-art YOLOX detection, followed by custom character-block recognition.

Each pipeline stage is tuned for reliability under low light, motion blur, and crowded environments.

Detection

Multi-class vehicle awareness

Classifies cars, buses, trucks, and motorcycles while estimating motion vectors for robust tracking.

Localization

Precision plate framing

YOLO anchors tuned to local plate formats ensure crisp crops for downstream OCR.

Recognition

Custom OCR ensemble

A lightweight classifier trained on multilanguage datasets keeps recognition in-spec even off-network.

Persistence

Intelligent tracking

A tracking-aware database aligns vehicles to detections, enabling analytics-ready history.

Live capture

Screenshots and flow

Inspect the interface, detections, and the end-to-end logic that keeps every plate readable.

License Plate Detection

Sharp bounding boxes isolate plates before OCR flags anomalies in real time.

Character Recognition

Colored overlays reveal confidence per character with audit trails for compliance.

Stack

Technologies powering Shahin

From inference to UX, every layer is hardened for edge-grade production.

YOLOv11 (Object detection) Python (Inference, orchestration) Rust (Licensing & IP safety) OpenCV (Video filters & metrics) PyTorch & TensorFlow (Training) SQLite + Aurora (Historical data) ReactJS (HQ dashboards) Docker (Deployment)