Color vision deficiency (CVD)

The color vision deficiency dataset is designed to train and evaluate a deep neural network for image recoloring aimed at compensating for color vision deficiencies, specifically protanopia and deuteranopia. The dataset contains two types of images:

Natural Scene Images: These are photographs of real-world scenes, filtered to include images with colors that are confusing or difficult to distinguish for individuals with color vision deficiencies.

Artificial Images: These are generated images that feature color pairs specifically known to be confusing for those with protanopia and deuteranopia. The images are created by selecting color pairs from the RGB space and generating patterns designed to highlight these confusing colors.

This dataset is used to train a neural network, employing the Swin transformer architecture, to enhance contrast and maintain naturalness in images for individuals with these types of color vision deficiencies