"Deep Captcha"

"The DeepCaptcha dataset is a large-scale synthetic CAPTCHA dataset designed to evaluate and benchmark optical character recognition (OCR) systems and AI-based text recognition models under varying levels of visual distortion and adversarial complexity. The dataset contains multi-level AI resistance samples generated using a configurable CAPTCHA engine that introduces controlled variations in character deformation, noise injection, background interference, color distribution, and text length diversity.DeepCaptcha provides structured metadata including resistance levels, character frequency distributions, text-length variability, and image dimension statistics, enabling systematic evaluation of model robustness across increasing difficulty levels. The dataset supports research in adversarial machine vision, AI security, automated bot detection, and robustness benchmarking of deep learning models.This dataset is intended for academic research, robustness testing, and the development of secure human-verification systems."