LPR Accuracy Whitepaper

iApp LPR — current vs new generation (measured)

Upgrading iApp License Plate Recognition

iApp LPR's current generation handles plate OCR well, and it is kept unchanged. Its make/model component, however, predates the recent Thai-market and EV wave — so it cannot recognize many current Thai or Chinese-EV vehicles. The new generation replaces make/model with a ConvNeXt-Small classifier trained on current Thai cars, and adds EV / Hybrid / Petrol inference and phased neighbor-country support — all backward-compatible.

NEW make/model top-1 (held-out val)
NEW top-5
iApp LPR (current) make accuracy
held-out test images / classes

Head-to-head: make/model on identical images

Both engines scored on the same held-out validation split (15% per class, seed 42 — never trained on). Current = iApp LPR (current generation); New = iApp LPR (new ConvNeXt classifier).

MetriciApp LPR (current)iApp LPR (new)
loading…
Loading measured results…

Training progression (NEW)

Iterative improvement as in-domain data deepened (images scraped via a clean-IP server, merged + de-duplicated):

ModelBackboneClassestop-1top-5
v1ConvNeXt-Tiny3595.05%97.69%
v2ConvNeXt-Tiny6497.84%99.41%
v3ConvNeXt-Small6498.24%99.61%
v5ConvNeXt-Small7397.44%99.49%
v6 (deployed)ConvNeXt-Small7399.18%99.79%

Capabilities added

Methodology & honest limitations

Generated from whitepaper/results.json (measured) + runs/thaicar_v6/report.txt.