CAPTCHA_CHAOS_79 - A Dataset of 79 Stylized Characters Tailored for Secure CAPTCHA Applications

Published: 22 April 2025| Version 2 | DOI: 10.17632/f9m6c45kb5.2
Contributors:
Tanmay Deshpande, Prajwal Chopade, Atharv Kokate,

Description

This dataset comprises a collection of approximately 9,000 high-resolution images representing 79 unique characters, including uppercase and lowercase alphabets (A–Z, a–z), numerical digits (0–9), and a variety of special symbols. Each character is rendered in multiple font styles, background textures, colors, and orientations, creating significant visual diversity and complexity. The dataset emphasizes variations in typography, contrast levels, noise patterns, and distortion effects, making it suitable for tasks involving character recognition, pattern analysis, and robustness testing against varied visual inputs. All images are uniformly labeled and categorized based on their respective character classes, ensuring ease of use for training and evaluation in machine learning workflows.

Files

Institutions

  • Vishwakarma Institute of Information Technology

Categories

Image Processing, Optical Character Recognition, Machine Learning, Character Extraction, CAPTCHA, Adversarial Machine Learning

Licence