This module runs on the JeVois-Pro NPU using a quantized deep neural network. It is derived from https://github.com/opencv/opencv_zoo/tree/master/models/license_plate_detection_yunet See LICENSE for license information.
Please note that the model is trained with Chinese license plates, so the detection results of other license plates with this model may be limited. See the screenshots of this module for examples, or search for "Chinese license plate" on the web for more images.
This module is mainly intended as a tutorial for how to run quantized int8 DNNs on the NPU using OpenCV and TIM-VX, here achieved through only small modifications of code from https://github.com/opencv/opencv_zoo - in particular, the core class for this model, LPD_YuNet, was not modified at all, and only the demo code was edited to use the JeVois GUI Helper for fast OpenGL drawing as opposed to slow drawings into OpenCV images.