I collected the data from paper titled "Holistic Recognition of Low Quality License Plates by CNN using Track Annotated Data [IWT4S-AVSS 2017]" which is created by Jakub Špaňhel, Jakub Sochor, Roman Juránek, Adam Herout, Lukáš Maršík, Pavel Zemčík under 2017 14th IEEE International Conference on Advanced Video and Signal Based Surveillance (AVSS).
For prediction, upload images from Validation_images folder.
demo link : here