This is a FasterRCNN detector trained in Matlab using 258 car plate images.
The neural network is Googlenet，trained in a gtx 950 in Matlab R2018b for about 70 minutes.
Learn more info at Train a Faster R-CNN deep learning object detector
https://ww2.mathworks.cn/help/vision/ref/trainfasterrcnnobjectdetector.html?#bvkk009-1-trainingData
How to use:
1.load plate_detector_googlenet_258imgs.mat into work space.
2.open plate_detection.m in matlab.
3.change line 4 " img = imread('6.jpg'); ",change the image to what ever car plate image you like.＜/br＞
4.click run,wait for around 30 seconds.
Issues：
1.Maybe need Matlab R2018b version or higher.
2.You may confront “Out of memory” , when detecting car plate in a video card that has 2GB of memory or less . If this problems occur , use cpu for detection.
3.The size of image must be larger than [224 224]. The minimum size is defined by the network's image
input layer.
Results:
