License Plate Detection
What it does. Where it can be used: camera mount height, distance, and angle of view.
Book a Demo
In Internal Testing
What it does:
The license plate detection model looks for a license plate when a vehicle is detected in video feeds.
Allows you to visually browse events by license plate. Allows you to filter car events by the presence of a license plate. Allows you to run and validate the chained license plate recognition models. Does not database or understand the characters.
Best Practices:
Prerequisite:
Requires vehicle detection and its camera placement requirements. Only runs when a vehicle has been detected.
Additional Camera Placement Requirement:
License Plates must be at least 20x20 pixels in size, requiring the car to be substantially closer than the vehicle detector.
Difference between License Plate Detection and License Plate Recognition
License Plate Recognition attempts to understand the question "What are the characters (numbers and letters) of this license plate?" License plate detection attempt to answer the question "Is there a license plate in this image and where is it?" Our license plate recognition model only runs inside the area where the license plate detection has occurred. If you've used a license plate recognition model before you may have gotten the "How am I driving?" bumper sticker results rather than the license plate. Running the license plate recognition model inside the license plate detection area reduces this problem.
Example of both License Plate Detection and License Plate Recognition running at the same time:
In most cases, the purpose of running LPD is LPR. Below is an example of both models running simultaneously.
License Plate Detection without License Plate Recognition:
With survail you can run license plate detection even if you do not want to run license plate recognition. Although we have not seen examples of regulation that prevents certain entities of running LPR in the way that we do with facial recognition, we decided to implement a flexible architecture that allows you to run LPD without necessarily running LPR. So, you can browse license plates visually even if you don't use the OCR features of LPR.
Not all license plate detections can result in the license plate recognition system being able to accurately identifying the letters.
This model was made by NVIDIA.