Fusion　design　is　an　innovative　design　approach　in　product　color　design.　The　fusion　design　makes　the　color　matching　richer.　It　not　only　can　be　simultaneously　with　a　variety　of　styles　but　also　fully　express　the　design　concept.　However,　the　manual　completion　of　the　fusion　design　process　takes　time　and　effort.　With　the　development　of　artificial　intelligence,　some　designers　apply　intelligent　algorithms　to　fusion　design　and　significantly　improve　design　efficiency.　However,　most　design　methods　only　achieve　style　migration　from　the　source　domain　product　style　to　the　target　domain　product　in　one　direction,　leading　to　the　resulting　design　solution　losing　some　of　the　style　information　of　the　target　domain　product.　We　propose　a　color-style　fusion　design　method　based　on　generative　adversarial　networks.　Take　the　eyeshadow　color　design　as　an　example.　Firstly,　the　source　domain　art　painting　dataset　and　the　target　domain　eyeshadow　dataset　are　collected,　and　the　CycleGAN　extracts　the　color　styles　of　both　to　fuse　to　generate　an　eyeshadow　color　scheme　with　artistic　style.　The　experimental　results　show　that　the　proposed　method　can　meet　the　color　design　concept　of　color　cosmetics　and　retain　the　original　art　painting　color　style,　providing　a　reference　for　designers　and　a　new　idea　for　product　fusion　design.　©　2023　IEEE.