Our paper titled “Deep Learning Driven Plates Recognition System” has been presented and published on the 17th International Scientific Conference on Industrial Systems – IS’17 which took place on Andrevlje, hotel Ceptor, October 4-6th.
Authors of this paper propose a new approach of developing the deep learning based method for plate recognition system. Proposed method consists of combining the state-of-the-art deep neural networks for both of the essential tasks: plate detection and character recognition which were trained and validated on artificially generated vehicle plates’ dataset.
Complete development process is described in details and a test system is implemented, as well. The proposed system is tested in the real-time environment and achieved overall accuracy is 98% for plate detection, 97% for letters recognition and 94.5% for digit recognition. The proposed system has almost real-time response and it is suitable for image and video based recognitions.
Full paper can be downloaded clicking this link.