Intelligent Clothing Fabric Color Matching Based On Popular Colors And Sentiment Analysis

With the development of color economy,competition in the clothing market has become increasingly fierce,and market-based demands have become more diversified.In the field of fabric color matching.On one hand,popular colors are increasingly valued by clothing brands and designers.When designers make color palettes for clothing fabrics with reference to popular colors,they not only pursue color matching innovations to attract consumers,but also consider maintaining the original style characteristics of clothing brands to build consumer loyalty.On the other hand,designers begin to pay attention to the emotional needs of users(consumers).They expect to echo the user's sentiment and the one conveyed by the color matching in the palette,and then formulate a suitable palette and complete the color matching of clothing fabrics.Therefore,this article proposes an intelligent clothing fabric color matching method with reference to popular colors and sentiment analysis,and provides designers with optional color matching schemes from these two perspectives.The specific work is as follows:(1)Color Palette Generation Method with Referring to Popular Colors: First,a clothing fabric style classification model based on Support Vector Machine(SVM)is designed to judge the style of original clothing fabrics,say,the clothing brand style;secondly,a popular color selection model based on discrete particle swarm optimization algorithm is proposed to generate color palette with referring to popular colors.In the algorithms,the fitness functions,the constraints and the particle correction strategies are formulated according to the matching principle of popular colors.The experimental results show that the proposed Color Palette Generation Method with Referring to Popular Colors can select the innovative colors from the current season's popular colors and form a palette with maintaining the original style of the clothing brand.The palette could subsequently provide a basis for the application of Clothing Fabric-Palette to Colors Networks(CF-PCN).(2)Sentiment to Palette Networks(SPN): The SPN network model,including generation network and discriminant network,is designed based on the idea of conditional generative adversarial networks.The generation network adopts a encoder-decoder structure,which is composed of a Bi-GRU(Bidirectional-Gated Recurrent Unit)-based encoding module,a condition enhancement module,and a GRU decoding module with an attention mechanism.Among them,the Bi-GRU coding module is used to analyze the context information in the multi-level(including words,phrases,and sentences)of the user's emotional needs.The condition augment module can make the generated palette more diverse,The GRU decoding module with an attention mechanism can focus on the key sentiment information in the context information and decode it into a palette.The discriminant network is used to judge true and false of the both generated color palette and the ground truth one.Compared with Heer and Stone's model,experimental results and user research show that the SPN model proposed in this paper can generate multiple palettes that better fit the user's emotional needs when multi-level user emotional needs are input.Therefore,the proposed model is more favored by users.The palette generated by the model in hence provides a basis for the subsequent application of the CF-PCN.(3)Clothing Fabric-Palette Color Matching Networks CF-PCN: The CF-PCN network model designed based on the idea of conditional generative adversarial networks includes generation network and discriminant network.Among them,the generation network is composed of the main color matching network and the conditional network.The main color matching network adopts the U-Net architecture,which uses jump connections to avoid feature loss during the down-sampling process;the conditional network integrates the palette color features into the main color through splicing.In the color matching network,the color information of the palette can be fully contained in the color matching result.The discriminant network is used to judge true and false of the both generated clothing fabric image and the ground truth one.The experimental results show that the color scheme generated by the CF-PCN model is 29.6% higher in the evaluation index palette color ratio than the result of Chang's model,indicating that the color matching results contain more palettes color information and the color matching quality is higher.When the color palette network is used to match the colors of clothing fabrics on the condition of the popular color palette and the sentiment analysis palette,the former can make the clothing fabrics after the color matching maintain the original style characteristics of the clothing brand and have the color matching innovation.The latter can reflect multi-level user emotional needs in diversified color matching,and provide designers with optional clothing fabric color schemes.