<b>IMLP: Iranian Motorcycle License Plates</b>

The <b>Iranian Motorcycle License Plates (IMLP) dataset</b> includes <b>1089 different images of motorcycles</b>, containing in total <b>1191 Iranian motorcycle license plates</b>. These images were captured and collected from the cities of <b>Islamabad Gharb, Kermanshah, Karaj, and Tehran</b>, and some were also obtained from the <b>Instagram social network</b>. Initially, motorcycles were filmed, and then one or two frames were selected for the dataset. The final images were randomly chosen from over <b>50,000 frames</b>.The dataset was built to ensure <b>diversity in conditions</b>:Captured using different cameras (Samsung S23 Ultra, Xiaomi Note 10 5G, Xiaomi Note 11s, Samsung J6, and several other mobile phones, as well as social network images).Collected from <b>different angles, distances, and lighting conditions</b> (daytime and nighttime).For annotation and benchmarking, <b>deep learning networks YOLOv8, SSD, and Faster-RCNN</b> were employed. The motorcycle plates were identified in a <b>two-stage process</b>:<b>Detection</b> of the plate region.<b>Recognition (OCR)</b> of the digits.The dataset is organized into two main parts: <b>Detect</b> and <b>OCR</b>, each containing <b>train, test, and validation splits</b>. Every folder includes images and annotations compatible with <b>YOLOv8, SSD, and Faster-RCNN</b>. In addition, <b>implementation files</b> for these networks are provided to facilitate reproducibility.This dataset and the associated Python code are linked to the following peer-reviewed publication:<b>Article:</b> <i>Deep Learning Based System for Automatic Motorcycle License Plates Detection and Recognition</i> <b>Authors:</b> Abdolhossein Fathi, Babak Moradi, Iman Zarei, Afshin Shirbandi <b>Journal:</b> <i>Signal, Image and Video Processing (Springer)</i>, Volume 18, Pages 8869–8879, 2024 <b>DOI:</b> https://doi.org/10.1007/s11760-024-03514-5<br>In this article, the IMLP dataset was introduced and evaluated using YOLOv8, SSD, and Faster-RCNN. The results demonstrated that <b>YOLOv8 achieved the best performance</b>, with <b>98.5% accuracy in the detection stage</b> and <b>99% accuracy in the recognition stage</b>. The dataset and source code are made publicly available to support further research in <b>intelligent transportation systems (ITS)</b>, particularly in the context of <b>motorcycle license plate detection and recognition</b>.