This issue recommends a set of license plate recognition open source project – Light-LPR.
Light-LPR is an open source license plate recognition project that can run on embedded devices, mobile phones and ordinary x86 platforms, aiming to support license plate recognition in open scenarios, with a license plate character recognition accuracy of more than 99.95% and a comprehensive recognition accuracy of more than 99%, supporting all domestic license plate recognition.
The Light-LPR technology uses MTCNN to detect the license plate and the four corner points for accurate positioning, and carries out deviation correction, and finally carries out end-to-end identification of the license plate number, and uses MNN as the inference engine. It has the characteristics of simple structure, flexible deployment, and ADAPTS to various computing platforms.
Light-LPR supports license plates:
Note: Y is supported, – unknown, E limited
1080P image recognition benchmark performance:
x86 Linux installation instructions:
git clone https://github.com/lqian/light-LPRcd light-LPR && mkdir build && cd buildcmake ../make
ARM platform Linux installation instructions:
git clone https://github.com/lqian/light-LPRcd light-LPR && mkdir build && cd buildcmake ../ -DLIGHT_LPR_ARCH=armmake
give a typical example:
You can read more on your own.