Improving webtoon accessibility for color vision deficiency in South Korea using deep learning

The webtoon industry is experiencing significant growth in terms of its user base and revenue, rapidly expanding from the local market to international ones. However, despite this growth, its accessibility has not improved significantly, with most efforts focusing on providing more webtoon content within the platform. Not much is known about who is excluded from the webtoon industry. To address this, this research particularly focuses on color vision deficiency (CVD) accessibility of webtoon content. Adapting the design science research framework, this research followed the four steps of design science research: problem identification and motivation, development and demonstration, evaluation, and communication. First, it organized the CVD issues in South Korea. Then, it proposed a technical solution using a deep learning model. Specifically, this research used a Pix2Pix generative adversarial network-based recoloring algorithm and provided the translated images. We evaluated the solution by using colorblind filtering and in-depth interviews with individuals with CVD. This research provides a well-rounded accessibility study on the webtoon domain by identifying issues, proposing a deep learning solution, and evaluating that algorithm with the people with CVD. A total of seven CVD issues in webtoon contents have been identified. Deep learning-based image translation model has been developed to improve the accessibility of CVD people to webtoon content. This study contributes to both CVD and media accessibility by structurally organizing the sequence of problem solving for accessibility issues in the webtoon domain.