Indian Number Plate Character Dataset (IURS) - ANPRINR
Description
The ANPRIND Character-Detection dataset is a curated collection designed for character-level labeling in Indian vehicle number plate detection. This dataset provides a valuable resource for training and evaluating computer vision models focused on Automatic Number Plate Recognition (ANPR) at the character detection stage. It contains a total of 8,168 images, with characters annotated in COCO format using Roboflow’s annotation tools to ensure accuracy and consistency. To prepare the dataset, each image underwent standardized pre-processing, including auto-orientation of pixel data with EXIF stripping, resizing to a uniform 640x640 resolution, and conversion to grayscale (CRT phosphor) to enhance clarity for character recognition tasks. To further improve dataset diversity and robustness, three augmented versions of each source image were generated. The augmentation techniques included horizontal and vertical flips with a 50% probability, equal probability of 90-degree rotations (none, clockwise, counter-clockwise, upside-down), random rotations between -15° and +15°, and random Gaussian blur ranging from 0 to 2.5 pixels. These augmentations simulate real-world variability and strengthen model generalization. The dataset is particularly suited for developing character-level ANPR systems tailored to Indian number plates, supporting applications such as traffic law enforcement, vehicle tracking, and smart city initiatives. By leveraging Roboflow’s annotation and export capabilities, the dataset ensures reproducibility and compatibility with modern computer vision pipelines. Researchers can also integrate it seamlessly into training workflows and extend its use for academic and applied projects in intelligent transportation systems. In summary, ANPRIND Character-Detection offers a high-quality, well-annotated dataset with diverse augmentations, making it a strong foundation for advancing research and applications in Indian vehicle number plate character detection.
Files
Steps to reproduce
The ANPRIND Character-Detection dataset was created for research in Indian vehicle number plate character 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 character 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 character detection tasks.
Institutions
- Kadi Sarva VishwavidyalayaGujarat, Gandhinagar