Adaptive Color Matching via CNN and GRU Architectures with Optimization through Adaptive Cuckoo Algorithm

Abstract

With the increasing influence of color on various fields, the current color matching technology that relies on manual color matching can no longer fulfill the color matching requirements of various fields. To meet the standardized and rational classification of colors and enhance the ability to match complex colors, an intelligent color matching technology based on convolutional neural network deep learning model is proposed. The research employs multi-layer convolutional networks to extract color space features, which are then combined with a gated recurrent unit neural network for modeling. The model captures color dependencies through gate updates and reset operations, followed by optimization via an adaptive cuckoo algorithm. Parameters are updated through Levy flight dynamics and dynamic elimination rules, enabling adaptive adjustments of step sizes and elimination probabilities with iteration counts. The model is ultimately applied to the DeepFashion dataset. During experiments, input data from the DeepFashion dataset were processed through feature layers using multidimensional tensors, ultimately generating RGB color schemes. The experimental results show that after 20 iterations of the model, the accuracy of the color matching method remains stable at around 93%, with a color difference range of 0.08-0.13 in warm tones and 0.03-0.06 in cool tones. In practical intelligent color matching, when the sample size of the color matching is increased to 10cm, the generation time of the color matching result stabilizes at around 2.8s, and the brand compatibility of the color matching result attains a 98% accuracy. The aforementioned results demonstrate that the intelligent color matching technology proposed in this research exhibits both high efficiency and accuracy. It effectively addresses the issues of sluggish color matching speeds and significant result errors prevalent in current color matching technologies, thereby enhancing the practicality of intelligent color matching solutions.