Peker, Murat2024-11-072024-11-072019978-1-5386-8086-5https://hdl.handle.net/11480/137721st Global Power, Energy and Communication Conference (IEEE GPECOM) -- JUN 12-15, 2019 -- Nevsehir, TURKEYIn this work, the object detection networks of TensorFlow framework are trained and tested for the automatic license plate localization task. Firstly, a new dataset is prepared for Turkish license plates. The images in the dataset are labeled with two classes which are the car and the license plate. Four different object detection networks were configured to run on Google's Colab environment. These network configurations were the Single Shot MultiBox Detector (SSD) using MobileNet features and Resnet50 features, the Faster Region Convolutional Neural Network (Faster R-CNN) using Inception layers for features, and the Region-based Fully Convolutional Networks (R-FCN) with Resnet101 features. These networks were compared to determine the performance of license plate localization. Different types of input images were used to test the algorithms.eninfo:eu-repo/semantics/closedAccessSSDFaster R-CNNR-FCNobject detectionlicense plate localizationConference Object1011052-s2.0-85070612373N/AWOS:000851517900020N/A