You signed in with another tab or window.Reload to refresh your session.You signed out in another tab or window.Reload to refresh your session.You switched accounts on another tab or window.Reload to refresh your session.Dismiss alert
A tool for generating Chinese license plate dataset for plate detecting When you are working on a Automatic Number Plate Recognition(ANPR) project, you may need thoundands of samples to train. Heres is my solution for generating chinese licence plate samples.
First Stage: Generate the plate
There are two kind of plates:1 real plate are cutted from pictures
2 Fake plate are formed with several random characters
Second Stage: Generate world(background)
I named background with ‘world’ in my code.There are two kinds of world image:1 image with only one colorI used only blue image for blue plate, yellow and white image could be added.2 Random image from some big datasets
Third Stage: Add objects to world
I add some negative objects for hard nagative mining, and adds only one plate to one world. I put the position of the plate relative to the world in the image file name. You can read the position from the image file name as ground truth. ##PrerequisitesI used Python 3.5.2 (from Anaconda 4.2.0)python-opencv 3.1
Here are some examples:
About
A tool for generating Chinese license plate dataset for plate detecting