Read every plate, even in haze.
PlateVision recognises vehicle license plates in complex environments using image enhancement and lightweight convolutional neural networks that need little training data and little computing power.
The problem
Standard plate readers fail when conditions get tough.
Haze, glare, low light and cluttered backgrounds reduce the accuracy of conventional recognition systems, especially when data and hardware are limited. That affects parking payments, access control and traffic safety.
F1 score for plate localisation in complex scenes.
F1 improvement over conventional image-processing methods.
F1 advantage over YOLOv3 in our tests.
Source: the founding research paper (IEEE Transactions on Intelligent Transportation Systems, 2025).
How it works
How it works
A clear path from data to decision.
Capture
Frames come from existing gate, parking or roadside cameras.
Enhance
ACE enhancement and dark-channel dehazing clean up difficult images.
Locate
The plate region is found using grayscale analysis and colour key elements.
Read
A lightweight CNN recognises the characters for logging or access decisions.
Product
What makes it different.
Every feature is grounded in the founding research.
Few-shot learning
Performs well even when little labelled training data is available.
Lightweight
Low computational cost, suited to edge devices and existing cameras.
Dehazing built in
Handles fog, dust and poor lighting with image enhancement.
Beats popular detectors
Outperformed SSD, YOLO and YOLOv3 by 25.62%, 27.16% and 18.20% in F1.
Who it's for
Who it's for.
Parking operators
Automatic entry, exit and payment.
Residential & corporate gates
Hands-free vehicle access control.
Traffic authorities
Enforcement and monitoring in real-world conditions.
The research
Built on peer-reviewed research.
- Few-shot lightweight license plate recognition using image processing and CNNs.
- Preprocessing with ACE enhancement and dark-channel dehazing for complex scenes.
- Localisation accuracy 85.95%, recall 83.24%, F1 85.14%.
- F1 improved by 48.26% over conventional methods and exceeded SSD, YOLO and YOLOv3.
Gunda, N. K., et al. (2025). Few-Shot Lightweight License Plate Recognition in Complex Environments for Intelligent Transportation Systems. IEEE Transactions on Intelligent Transportation Systems.
Team
The team behind it.
Founded in 2025 by a student team from the faculty below.
Faculty of Engineering Technologies (Computer Technology Engineering)
Computer vision, embedded deployment and system integration.
FAQ
Questions we hear often.
Does it need new cameras?
No. It is designed to work with standard existing cameras.
Can it run without the cloud?
Yes. Its lightweight design suits on-site edge devices.
Is this a finished commercial product?
It is a research-based student startup moving from validated prototype toward pilots with real users.
Can we partner or run a pilot?
Yes. We are looking for organisations willing to test the solution in real conditions and share feedback.
Bring PlateVision to your organisation.
We are onboarding pilot partners now. Read the research and get in touch.