論文研究關鍵字

首頁Home › 人工智慧與新興科技 AI & Emerging Tech

NO.825

風格轉換

Style Transfer

1這是什麼What is it?

風格轉換(style transfer)將一張圖像的內容與另一張圖像的風格重新組合,Gatys 等人(2015)以卷積神經網路的特徵統計分離並重組內容與風格,開啟神經風格轉換。後續發展出快速前饋網路、任意風格轉換與 CycleGAN;其問題涉及「風格」的可定義性與對藝術家風格的挪用。

Style transfer recombines the content of one image with the style of another, and Gatys et al. (2015) used convolutional network feature statistics to separate and recombine content and style, opening neural style transfer. Later developments include fast feed-forward networks, arbitrary style transfer and CycleGAN, and its issues involve the definability of "style" and appropriation of artists' styles.

2常被用在哪種研究問題Typical research questions

設計與評估型Design & evaluation分析/歸納型Analytical

常見問法:風格轉換結果的視覺品質如何評價?

示範寫法Example

風格轉換技術重塑圖像風格。本研究比較不同風格轉換結果,並以觀者評價檢驗其美感與風格保真度。

Style transfer techniques reshape image style. This study compares results of different style transfers and tests their aesthetics and style fidelity through viewer evaluation.

撰寫內容,非真實研究結果,請與指導教授討論後改寫。

Written as a model example, not a real finding. Please revise with your advisor.

3怎麼測它How to measure it

觀者評分+專家評審

Measurement notes are written in Chinese. Switch to 中/EN to read them.

4代表文獻Key references

英文論文(2000 年後)English paper (post-2000)

Gatys, L. A., Ecker, A. S., & Bethge, M. (2016). Image style transfer using convolutional neural networks. CVPR 2016.

中文論文(2000 年後)

中文論文為預填檢索連結,請自行挑選一篇 2000 年後的期刊論文並核對作者、年份、出處後再引用。

英文引用來自模型整理,引用前請務必核對並閱讀原文。English citations are compiled by a model. Verify and read the original before citing.

‹ 創意程式設計Creative Coding虛擬實境 ›Virtual Reality