ParkVision 06 / 25
Vehicle-type classification & recognition API
Classifies the vehicle by UNECE category and serves everything through a REST API and web UI.
Problem
Barrier and parking systems need not only the plate but the vehicle class, in a format they can integrate.
What it does
- Classifies car, SUV, van/pickup, trucks and buses (UNECE)
- Uses a plate → vehicle-type registry to raise confidence
- REST API: image in, JSON out (plate, plate box, confidence, type, processing time)
- Web UI for demos and visual checks; health endpoint; edge capture on Raspberry Pi
How it works
- Detection box + frame → neural vehicle-type classifier (runs in parallel with OCR)
- Registry lookup by recognized plate
- Result assembled into JSON for barriers, parking and accounting systems over HTTP(S)
Stack & models
PythonPyTorchOpenCVONNX RuntimeHTML/JSGo & C helpersLinux / Windows x86-64
Numbers
- Server: 8 GB RAM, 4 GB VRAM for GPU mode; images up to 45 MB
Figures come from project docs/README; no other benchmarks are claimed.