An Intelligent Big Data Approach to Artistic Color Scheme Generation Using StyleGAN2-ADA and Spark Framework

Existing color scheme generation often relies on rule-based algorithms, with a single style and difficult to apply to large-scale color data. This study applies StyleGAN2-ADA under the Spark framework and proposes an intelligent color scheme generation method in big data scenarios: the color samples of the dataset are integrated and cleaned, and the RGB values are converted to the CIE Lab color space. Then, the generator and discriminator in StyleGAN2 are used together on the image samples, and the discriminator training strategy is dynamically adjusted in combination with adaptive discriminator augmentation (ADA). By controlling the input noise vector, feature vector parameters and diversified style templates of StyleGAN2-ADA, diversified art style color schemes are generated. The Spark framework is constructed, and the model parameter update is accelerated through distributed training methods. The training effect is monitored in real time in combination with TensorBoard, and large-scale data are processed quickly through data partitioning and task scheduling. The research results show that the diversity of color schemes is 27.9% higher than that of the classic color palette model, and the average style matching degree reaches 87.7%. This method significantly improves the efficiency and quality of color scheme generation, and provides a new solution for color matching in art design.