Perform OCR on a single image or multiple images in a folder to get the License Plate Number. Input image required is that of a cropped license plate. Example images are provided in license_plate folder.
Character segmentation is performed first and then the recognition model is applied to the individual characters.
Keras sequential model is used to perform OCR. Input size of each character is 28x28. Label file is provided for all the characters to recognise.
For ease of use a config file is provided for model path, label path, image path, folder path. if is_folder = 0, then, OCR is performed on single image. If is_folder = 1, then, OCR is performed on a collection of images present in a folder.
For training your own model train_alpr.py is provided. Data is provided in ocr_char. Provision for augmentation also given.