As the name suggests, this code guesses an area of region that might be the number plate of a car. The accuarcy of the code relies a bit on the lighting in the input image.
The detection is two fold. First, the code detects the region of image in which the car is present. This bit is achieved using pretrained network and code based on YOLOv3 network. A lot of my code also draws inspiration from this amazing blog.
Then, the code attempts to guess the region of image where the number plate may be present. I have expanded on the idea I read about and explored in this repository.
P.S. The pretrained weights for YOLOv3 is too big to upload, so I am provideing the relevant link here
Step 1: Clone the Repository:
$ git clone https://github.com/Mastermind0100/Automatic-Number-Plate-Recognition-System.gitSave the image you want to use for detection in the same directory as testimage.jpg (or whatever extention you are comfortable with).
Step 2: Open cmd in the same directory (or open cmd and go to the directory where all the files are saved)
Step 3: Type the following command:
python new_yolo.py --image testimage.jpg --config yolov3.cfg --weights yolov3.weights --classes yolov3.txtNote, the testimage.jpg is to be replaced by the image you want to test
Step 4: Hope you get the output you want!
P.S. Press any key to obtain further images :)
You can use this code in conjunction with my OCR (Optical Character Recognition) repository to obtain a full fledged ANPR (Automatic Number Plate Recognition) system! :D