active active

acceleration

The OpenALPR commercial engine supports execution on NVIDIA hardware to enable high-speed, real-time License Plate Recognition. If your application uses mobile cameras or captures vehicles moving in excess of 60 miles per hour, OpenALPR Nvidia GPU acceleration makes it possible to deliver real-time performance and state-of-the-art accuracy.

Highest Possible Accuracy. OpenALPR measures license plates more confidently when it sees the same plate consecutively. With Nvidia GPUs, OpenALPR is able to form highly confident results because it can analyze every frame of the input video stream.

Lower overall cost. Utilize fewer, high-resolution cameras and capture multiple lanes of traffic with a single system. Using Nvidia hardware, OpenALPR is able to monitor 4k video (3840x2160 pixels) across all 4 lanes of a highway with a single camera.

More Headroom for your Application. License Plate Recognition is a computationally intensive operation. By moving the processing off of the CPU and onto the GPU, you have more resources available for your own services. Keeping the CPU usage light ensures that your application runs smoothly and you deliver a seamless experience for your end-users.

Power and Space Efficient. OpenALPR running on Nvidia hardware recognizes many more plates per watt of power compared to CPU processing. A single Nvidia GPU processes many simultaneous video streams in the data center. Similarly, the low-power Nvidia Jetson device performs real-time license plates recognition in the field. OpenALPR supports the energy-efficient Nvidia Jetson platform, as well as desktop and server-class GPUs, such as the Tesla, Quadro, and GTX product lines.

OpenALPR Nvidia acceleration is available today on Ubuntu Linux. Contact us for an evaluation.

Watch OpenALPR recognize license plates from a high-resolution (1080p) video stream in real-time. This demo is running on a Nvidia Tesla P4 at greater than 50 frames per second processing speed. You may have to slow down the playback to read the plate numbers, the cars are moving so quickly that it's hard for a human eye to keep up!


Benchmarks

There are a variety of factors that influence how many frames per second are sufficient for ALPR. It is dependent on the angle of capture, resolution, speed of the vehicles, and other factors. A good rule of thumb is that 5fps of processing per camera stream will allow for low-speed vehicle capture, and 10fps or greater is sufficient for high speed.

Performance benchmarks are detailed below. The Jetson platform is an ultra-low power, embedded device that is a great solution for in-vehicle or in-camera capture. The Tesla P4 is installed in server-class hardware and is powerful enough to process multiple video streams simultaneously.

System FPS @ 1080p FPS @ 720p
Jetson TX-2 (Embedded) 5.5 10.5
Tesla P4 (Server) 50.5 92.5

Metropolis

OpenALPR is an official partner with Nvidia on the Metropolis intelligent video analytics platform.

Nvidia Metropolis™ is bringing advanced AI to cities around the world. OpenALPR is proud to be selected as the license plate recognition engine to support this exciting initiative.