Sapta Lipi HCR : Handwritten Character Dataset for Seven Indic Scripts

Description

Indic scripts have played a significant role in preserving the linguistic, literary, cultural, and historical heritage of the Indian subcontinent. Several Indian languages have evolved through distinct writing systems, each possessing unique character structures, writing styles, and graphical characteristics. These scripts are widely used in historical manuscripts, literary works, inscriptions, and other forms of written documentation. The development of automated recognition systems for Indic scripts is an important research area, particularly for handwritten character recognition, digital document preservation, and multilingual optical character recognition (OCR). This dataset presents a collection of handwritten character images from seven Indic scripts: **Devanagari, Kannada, Bengali, Nandinagari, Sharada, Telugu, and Malayalam**. It is designed to support research and experimentation in multiscript character recognition, handwritten character classification, and image-based script processing. The dataset is organized into separate categories for vowels, consonants, and numerals. Each character class contains 100 handwritten image samples. All images are converted into grayscale-derived binary (black-and-white) representations and normalized to a uniform resolution of **28 × 28 pixels**. This standardized image format facilitates consistent preprocessing and provides a suitable input representation for machine learning and deep learning-based character recognition models. For the seven scripts with the currently documented class counts, the dataset comprises 426 distinct character classes and 42,600 handwritten character images. The class distribution includes consonants, vowels, and numerals, providing a structured and balanced collection for character-level recognition experiments. The dataset can be utilized for developing and evaluating machine learning and deep learning approaches for handwritten Indic character recognition, multiscript classification, comparative analysis of script recognition models, and preprocessing techniques for low-resolution handwritten character images. It may also serve as a foundation for further research in multilingual OCR systems and digital preservation of Indic language resources.

Institutions

• VIT-AP University

  Andhra Pradesh, Guntur

Categories

Licence