Yes, it works. Areatec's Aretron was trained specifically to operate in the adverse conditions of the real Brazilian world — not just in controlled laboratory scenarios [1].
Performance by Adverse Condition
| Condition | Read Rate | How the AI Handles It |
|---|---|---|
| Sun glare (backlight) | 97%+ | Automatic HDR + multiple simultaneous exposures |
| Heavy rain | 95%+ | Raindrop removal algorithm + infrared lighting |
| Dirty plate (mud/dust) | 93%+ | Focal Loss for partially obstructed characters |
| Dented plate | 90%+ | AI geometric reconstruction |
| Nighttime (no lighting) | 98%+ | Dedicated infrared lighting |
| High speed (180 km/h) | 96%+ | Ultra-fast shutter + Edge AI processing |
What Sets the Focal Loss Algorithm Apart
Focal Loss is a deep learning technique that gives more weight to difficult cases during training. While conventional algorithms "give up" on partially illegible plates, Aretron AI was trained on millions of images of plates in real adverse conditions collected on Brazilian streets [1].
Processing happens locally inside the vehicle (Edge AI), without relying on an internet connection. This ensures continuous operation even in areas with weak or no signal.
Legal Validity
For a reading to carry legal validity, the system must reach a minimum confidence of 95% in the identification. Readings below that threshold are discarded automatically and do not generate tickets, protecting the driver against false positives [2].