GAN-driven color scheme generation model for brand identity design

Abstract

In the realm of visual communication and marketing, color plays a crucial role in shaping brand identity and influencing consumer perception. However, traditional color selection methods are often subjective, inconsistent, and lack scalability, making it difficult to maintain coherence across diverse branding applications. To address this challenge, this research proposes a Generative Adversarial Network (GAN)-driven Color Scheme Generation Model that automatically produces brand-consistent color schemes tailored to specific brand values and emotional tones. The approach begins with the collection of branded images and logos from a wide range of industries, each annotated with associated brand attributes and emotional descriptors. During the pre-processing stage, images are converted into YCbCr and RGBY color spaces to preserve both luminance and chromatic information. The images are then resized to a uniform resolution and normalized for model input. Feature extraction is conducted using a pre-trained Super-Resolution Convolutional Neural Network (SRCNN) to capture fine-grained visual patterns relevant to branding aesthetics. These feature maps are then input into a Seven-spot Ladybird Optimized Palette Generative Adversarial Network (SLO-PaletteGAN) model. The generator creates color schemes, and the discriminator is trained using style-aware loss functions and semantic label constraints to ensure visual coherence and brand relevance. To further refine the generated palettes, a reinforcement learning-based chroma fine-tuning module is integrated, enhancing color accuracy and perceptual alignment. Quantitative evaluations using metrics such as Minimum Inter-theme differentiation (Inter-TD-min), Maximum Inter-theme differentiation (Inter-TD-max) of 1.9724, Intra-theme differentiation (Intra-TD), and Template bias (TB) demonstrate that the proposed model outperforms existing color palette generators. This research provides a scalable, intelligent, and data-driven solution for creative color scheme generation, advancing the capabilities of modern branding design systems.