透過您的圖書館登入
IP:3.135.246.193
  • 學位論文

探討使用生成對抗網路自動生成商標之品牌態度

Evaluating Brand Perception on Logo Synthesized by Generative Adversarial Network

指導教授 : 盧信銘

摘要


用人工智慧自動生成商標可以節省設計商標時的時間與金錢,並提供設計師與企業主管商標設計時的靈感。先前研究使用生成對抗網路自動生成商標,並使用定量指標衡量模型的表現。我們的研究除了用定量指標衡量模型表現,也希望用商標設計元素與品牌態度分析生成之商標,從一個管理的觀點衡量生成對抗網路模型之表現。我們使用兩種近期熱門的生成對抗網路之變異模型生成商標,比較兩者模型的表現,同時也與真實的商標做比較。我們選圖形特徵與生成對抗網路相似之真實商標進行比較。此研究結果顯示,兩種生成對抗網路生成商標之品牌態度也有顯著差異。另外,研究結果也顯示生成對抗網路與真實商標之品牌態度(尤其美觀與整體好感)有顯著差異,因此生成對抗網路所生成之商標較適合當作草稿,較不適合直接使用。不過也因為生成對抗網路所生成之商標較抽象,因此適合應用在各領域與產業的企業。此研究是第一個使用管理的觀點衡量生成對抗網路生成之商標。而我們的研究結果可以提供未來生成對抗網路應用與改進之方向,並提供設計師與企業主管關於生成對抗網路商標之參考。

並列摘要


Logo generation using Artificial Intelligence can save time and provide initial ideas for managers and designers. Prior studies use Generative Adversarial Networks as the model architecture and evaluates the model performance using quantitative metrics. In addition to the image quality, this study evaluates the GAN generated logos in terms of brand perception, affection and logo characteristics, with the aim to understand the brand performance of GAN generated logos. We compare logos generated from two GAN variation models and logos from the real-world that have similar logo characteristics as the GAN generated logos. Results show that there are significant differences in brand perception and affection (especially aesthetic appeal and liking) of real-world logos and GAN generated logos. We also find differences in brand perception and affection of the two GAN variation models. Overall, our research results suggest that GAN generated logos should only be used as initial drafts; however, because GAN generated logos are very abstract and the visual characteristics are less defined, they can be applied to various kinds of industries and companies. The flexibility of GAN generated logos could provide managers or designers inspiration for logo designs. This study is the first to conduct comprehensive evaluation on GAN generated logos. Our results assist designers and managers to better decide how to select and utilize GAN synthesized logos and provide managerial insights on GAN generated logos.

參考文獻


Atarsaikhan, G., Iwana, B. K., Uchida, S. (2018, 2018-04-01). Contained Neural Style Transfer for Decorated Logo Generation. Paper presented at the 2018 13th IAPR International Workshop on Document Analysis Systems (DAS).
Balmer, J. M. (1998). Corporate identity and the advent of corporate marketing. Journal of marketing management, 14(8), 963-996.
Balmer, J. M. T. (2001). Corporate identity, corporate branding and corporate marketing ‐ Seeing through the fog. European Journal of Marketing, 35(3/4), 248-291. doi:10.1108/03090560110694763
Bottomley, P. A., Doyle, J. R. (2006). The interactive effects of colors and products on perceptions of brand logo appropriateness. Marketing Theory, 6(1), 63-83. doi:10.1177/1470593106061263
Cian, L., Krishna, A., Elder, R. S. (2014). This Logo Moves Me: Dynamic Imagery from Static Images. Journal of Marketing Research, 51(2), 184-197. doi:10.1509/jmr.13.0023

延伸閱讀