Deep Learning-Based License Plate Recognition Software


Smart License Plate Recognition (Smart LPR) is a software library conveniently designed to work with any existing cameras and processors. It's the ideal cost-effective software choice for application developers who want to streamline their coordination efforts and contribute to building safer cities, leveraging advanced smart technologies.

Note: Images used of license plates are created and shown for demonstration purposes only.


State-of-the-art deep learning technology
Scan and capture accurate license plate data even in the most demanding of scenes—including low resolution and blurry images, poorly lit nighttime road conditions, and license plate protectors, including IR shields.
Real-time processing
A perfect pairing between the Sony video pipeline and GPU resource management optimization technique and LPR software leads to fast processing in edge devices. Smart LPR boasts a processing performance of 35fps (equipped with NVIDIA Jetson TX2).
Cost-effective AI solution
Take the first step to build a smarter, safer city. By removing unnecessary hardware from our sensing software, Smart LPR is a cost-efficient software library that provides an economical alternative for networks of mobile and stationary camera setups commonly seen in license plate tracking applications for law enforcement, Electronic Toll Collection (ETC), traffic agencies, parking lots and garages.
Hardware flexibility
Equipped to work with a variety of IP, CSI, and USB cameras as an input source, Sony LPR software is easy to install, use, and integrate into an organization's daily workflow. Currently supported by Linux OS and NVIDIA Jetson Series GPUs, Sony LPR can also be adapted to meet specific partner needs for vehicle tracking.
4K video resolution
Analyze video streams in up to 4K video resolution (UHD) with Sony LPR. Designed to identify license plate numbers in real-time, it analyzes video in ultra-high definition and outputs a text representation of US license plate characters with a high degree of certainty.
Added recognition for enhanced confidence
With 60 px W x 30 px H minimum pixels per foot per plate recognition, personnel can accurately review data with increased confidence. Additional data measures include plate image cropping, confidence level support, and state recognition result (partially supported).


Operating system
Linux Ubuntu 18
NVIDIA™ Jetson series (Jetson TX2, Nano, TX1, Xavier)
NVIDIA™ JetPack Software
Video input
RTCP (H.264, H.265, JPEG)
USB (RAW, H.264, H.265, JPEG)
File (H.264, H.265, JPEG, PNG)
For more OS/GPU options please contact your local Sony representative.