生成式人工智慧對於南雲治嘉配色認知描述之研究

A Study of Generative Artificial Intelligence in the Perception and Description of Nagumo Haruyoshi Color Image Chart

人類透過線條、形狀、紋理、空間及色彩形成圖像傳遞信息表達情感，這些視覺元素的組合，經過時代更迭，演化為設計基礎。設計便是策略性將其整合，用以影響觀眾的認知、情感與行動。其中，色彩的搭配與掌握，對於作品情感調性和說服力更具關鍵影響力。隨著科技發展，生成式人工智慧（Generative Artificial Intelligence,GAI）如今已有突破性的成果並被廣泛應用。然而，人類對色彩的認知融合了情感、文化背景和個體經驗，究竟GAI所生成的圖像能否有效表現人類對於色彩與色調的情感?其色彩表現的限制與局限為何?值得深入探討。本研究以DALL-E作為研究對象，從南雲治嘉配色圖表中挑選30個色彩主題，並以DALL-E生成之圖像做為研究樣本，透過深度訪談進行分析。研究顯示，雖然大部分生成圖像在配色的識別性和統一性均獲得認同，但在詮釋心理層面細膩情感主題時，仍存在挑戰需要人工介入。研究也強調結合深度學習與豐富文化數據的重要性，以提高GAI在設計及藝術創作中的應用價

Throughout history, humans have utilized lines, shapes, textures, spaces, and colors to craft images that convey information and express emotions. Over time, these visual elements have become the cornerstone of design, evolving through accumulated experience. Design is the strategic integration of these elements, serving as a practical means to communicate messages and concepts, while influencing the audience's perception, emotions, and actions. Among these elements, the use of color and its mastery are particularly crucial, as they significantly affect the emotional tone and persuasiveness of a design. With the advancement of technology, Generative Artificial Intelligence (GAI) has achieved significant breakthroughs and is now widely applied. However, human perception of color is influenced by emotions, cultural background, and personal experiences. Can GAIgenerated images effectively convey the subtle emotional nuances associated with color and tone? What limitations and constraints exist in its color representation? This study focuses on DALL-E as the subject of research, selecting 30 color themes from the Nagumo Haruyoshi Color Image Chart and analyzing the images generated by DALL-E through in-depth interview. The findings reveal that although most images are recognized for their color accuracy and unity, there are challenges in conveying subtle emotional themes, necessitating manual adjustment. The study also emphasizes the significance of integrating deep learning with extensive cultural data to enhance the application value and accuracy of Generative Artificial Intelligence in design and artistic creation.