Saudi License Plate Characters Dataset

Data Composition

Each image in this dataset is provided with two corresponding annotation files: an XML file and a TXT file formatted for YOLO (You Only Look Once) model training. These annotations include bounding boxes around the characters on the license plates, taking into account the paired nature of Arabic and Latin characters or Eastern and Western numerals. Bounding boxes ensure precise localization of the characters, making this dataset highly suitable for character-level recognition.

The images were sourced from a combination of publicly available data on the internet and original photographs taken by mobile phones. Each image was manually annotated to ensure high accuracy in labeling. The dataset covers a wide range of real-world scenarios, including varying lighting conditions, plate orientations, and image resolutions, making it versatile for robust OCR model development.

Applications

This dataset can be utilized for a wide array of applications:

• Automatic License Plate Recognition (ALPR): Enhancing the recognition of vehicle license plates in real-time applications such as traffic monitoring, toll collection, parking management, and law enforcement.
• Multilingual OCR: Developing and testing OCR systems that can handle multilingual characters, particularly those using Arabic script.
• Deep Learning Models: Training deep learning models for object detection, particularly in scenarios requiring precise recognition of small and complex character sets.
• Smart Cities and Surveillance Systems: Automating traffic management and surveillance systems by integrating license plate recognition for vehicle tracking.

File Formats

• Images: High-quality images in various formats (e.g., JPG or PNG) that capture license plates under different conditions.
• Annotations: XML and YOLO-formatted TXT files with detailed bounding boxes, providing structured data for easy integration with various machine learning frameworks.

Class Labels

The dataset includes 27-character classes, encompassing both Arabic and Latin characters, as well as numerals. This dual representation allows for flexibility in training models that can recognize characters in different formats across multiple languages.

Key Features

• 593 High-Quality Images: Covering diverse real-world conditions to ensure robustness and generalization of models.
• Multi-Class Annotations: 27 classes of license plate characters, supporting multilingual character detection.
• Bounding Boxes for Dual Characters: Carefully annotated bounding boxes for both Arabic and Latin representations of characters on license plates.
• Manually Labeled: All annotations are manually verified to ensure high accuracy.