Gujarat-HWD: A Dataset of Gujarati Handwritten Digits
Description
The Gujarat-HWD dataset offers a comprehensive collection of 11,000 images of handwritten digits in the Gujarati language, representing numerals from 0 to 9. This dataset is primarily designed for applications in optical character recognition (OCR), machine learning, and deep learning, with a particular focus on regional language processing. Gujarati, one of the most widely spoken languages in India, uses a script distinct from Devanagari. However, despite its extensive use, there is a notable lack of publicly available datasets for Gujarati handwritten digits. The Gujarat-HWD dataset bridges this gap by providing a clean, labelled, and diverse set of images that can aid researchers and developers in building effective recognition models for regional scripts. The dataset has been developed through a systematic process involving the collection, scanning, and preprocessing of handwritten digit samples provided by more than 350 individuals from various age groups and educational backgrounds. It is well-suited for training and evaluating classification models using standard convolutional neural network (CNN) architectures. Additionally, the dataset can be extended for use in cross-lingual digit recognition, handwriting analysis, and other regional OCR systems.
Files
Steps to reproduce
Download and Extract the Dataset Download the dataset from Mendeley Data. Extract it to your working directory. The dataset contains 10 folders named 0 to 9, each with 1,100 handwritten digit images. Install Required Python Libraries Load and Preprocess the Images Read images using OpenCV or PIL. Convert each image to grayscale (if not already). Resize each image to 128×128 pixels. Normalize pixel values to the range [0, 1] for model input.