7 Vehicle Classes Detection at 98.4% Accuracy

Vehicle Classification Without Extra Sensors

VEBZE's inference pipeline returns more than a plate string. Every API response includes a vehicle class label — derived entirely from the same camera frame used for plate recognition, with no lidar, radar, or loop detectors required.

The 7 Classes

Class	Typical Examples
Sedan	Passenger cars, taxis
SUV	Crossovers, 4×4s
Truck	Heavy goods vehicles, semi-trailers
Motorcycle	Bikes, scooters
Bus	City buses, coaches
Van	Panel vans, minivans
Pickup	Light trucks, utes

Overall accuracy: 98.4% across all classes in controlled evaluation.

How Classification Works

The classification head runs in parallel with the OCR head inside the same TensorRT graph — zero additional inference time. The model uses vehicle silhouette, roof profile, and wheelbase ratio to assign a class label with a confidence score:

{
  "plate": "34ABC123",
  "vehicle_class": "SUV",
  "vehicle_class_confidence": 0.97,
  "ocr_confidence": 0.99
}

Business Value by Class

Trucks and HGVs — trigger separate billing tiers in toll systems, restrict entry to height-limited zones, route to designated loading bays.

Motorcycles — exempt from certain access rules, apply different parking rates, flag in security zones.

Buses — grant bus-lane priority, coordinate with smart traffic signals, log for public transit analytics.

Accuracy by Class

Class	Accuracy
Sedan	99.1%
SUV	98.7%
Truck	98.9%
Motorcycle	97.2%
Bus	99.3%
Van	97.8%
Pickup	98.1%

Conclusion

Vehicle classification is included in every VEBZE API response at no extra cost — same frame, same latency, more structured data.