Indian Number Plate Dataset (IURS) - NPDS
Description
The NPDS dataset is a specialized collection created for the purpose of Indian vehicle number plate detection. This dataset provides a robust foundation for training and evaluating computer vision models in applications such as traffic monitoring, automated surveillance, and intelligent transportation systems. It contains a total of 19,309 images, with number plates annotated in COCO format using the Roboflow annotation tool, ensuring accuracy and consistency across the dataset. To prepare the images for model training, several pre-processing steps were applied, including auto-orientation of pixel data with EXIF stripping and resizing all images to 640x640 pixels. In addition, extensive augmentation techniques were used to increase dataset diversity and improve model generalization. These augmentations include horizontal and vertical flips with a 50% probability, random 90-degree rotations, random rotations between -15° and +15°, random shear transformations up to ±10° horizontally and vertically, and the application of salt-and-pepper noise to 1.29% of pixels. Each source image was expanded into three augmented versions, making the dataset highly versatile for real-world variability. The NPDS dataset is particularly valuable for developing Automatic Number Plate Recognition (ANPR) systems tailored to Indian vehicles, supporting use cases such as traffic law enforcement, vehicle tracking, and smart city initiatives. Its integration with the Roboflow platform allows seamless collaboration, dataset management, and export for training across multiple frameworks, while also enabling deployment-ready models with active learning support. In summary, NPDS is a carefully curated and annotated dataset designed to advance the field of Indian vehicle number plate detection, offering high-quality labeled data, diverse augmentations, and compatibility with modern computer vision pipelines through Roboflow’s ecosystem.
Files
Steps to reproduce
The NPDS dataset was created for research in Indian vehicle number plate detection using a combination of web scraping and manual image collection. Images were gathered from diverse online sources and manually curated to ensure relevance and quality. Once collected, the dataset was imported into the Roboflow platform, where all annotations were performed. Number plates were labeled in COCO format using Roboflow’s annotation tools, ensuring consistency and accuracy across the dataset. After labeling, the annotated dataset was exported directly from Roboflow, making it ready for training and evaluation in computer vision workflows. This straightforward process Data collection, annotation in Roboflow, and export provides a reproducible workflow that others can follow to build similar datasets for number plate detection tasks.
Institutions
- Kadi Sarva VishwavidyalayaGujarat, Gandhinagar