High Accuracy Data Management and Processing Models

Mallioris Panagiotis, Sidiropoulos Athanasios, Bechtsis Dimitrios, Vlachos Dimitrios

plate recognition, neural network, categorical cross entropy, TensorFlow, Keras

Data management and processing methods are used for efficiently handling time consuming and critical activities. With the up rise of digital technologies, data processing is used for supporting operational processes in everyday activities at a 24/7 perspective.  Data sources vary from simple numerical data and text data, to images and videos that should be efficiently analyzed in order to be used at the decision-making process. The proposed research work focuses on the implementation of data processing algorithms for high accuracy license plate recognition, that could be used in Digital Supply Chains. Open source frameworks and tools (TensorFlow and Keras API) were used, in order to implement neural network methods and examine their accuracy.  After a brief introduction, several methods are presented with the use of ssd_mobilenet_v1_coco and ssd_mobilenet_v1_ quantized TensorFlow models. The final model successfully recognizes license plates with high accuracy.

https://doi.org/10.31871/IJNTR.6.12.25