{"id":"deepdream","no":574,"name":"DeepDream","ja":"ディープドリーム","era":"2015–","family":"デジタル芸術","kind":"技法","collection":"Dictionary expansion 30","addedAt":"2026-08-22","source":{"title":"Google Research — Inceptionism: Going Deeper into Neural Networks","url":"https://www.research.google/blog/inceptionism-going-deeper-into-neural-networks/"},"reference":{"status":"verified","pageUrl":"https://www.research.google/blog/inceptionism-going-deeper-into-neural-networks/","credit":"Google Research — Inceptionism: Going Deeper into Neural Networks"},"essence":"画像認識ネットワークが学習した特徴を反復して増幅し、雲や岩や葉の中へ目、動物、塔のような形を過剰に見いださせるGoogle発の可視化技法。認識の誤読そのものが、反復する幻覚的な表面になる。","cues":["画面全体に反復する目と動物の顔","元の輪郭へ沿って増殖する渦状の質感","細部を見るほど別の像が現れる多尺度の反復","高彩度の青・黄・紫が連続してうねる"],"intents":["技術","未来","高揚"],"works":["機械が像をどう誤認するかを、一目で分かる展示や批評画像にするとき","音楽映像やジャケットで、元風景を残しながら幻覚的な多尺度の表面へ変えるとき"],"recipe":{"type":"文字は処理後に置き、細部の反復と競合しない単純な太字で短く保つ。","layout":"元画像の大きな構図を保ち、複数の縮尺で同じ特徴を増幅して奥へ続く反復を作る。","material":"学習済み画像認識モデルの層を選び、その活性を勾配上昇で反復して強める。"},"avoid":"単なる渦巻きフィルターや犬の顔の貼り込みではない。どの層の何が増幅され、元画像とどう結び付いたかを残す。","colors":["#21134A","#2FD5C4","#F0C84B"],"motif":"digital","pair":["generative-art","psychedelic"],"study":["feature visualization and layer selection","iterative zoom and octave scales","GoogLeNet ImageNet feature bias"],"en":{"essence":"Google's visualization technique repeatedly amplifies features learned by an image-recognition network, making eyes, animals and towers emerge excessively from clouds, rocks and leaves until misrecognition becomes a recursive hallucinatory surface.","cues":["Eyes and animal faces repeated across the frame","Swirling texture multiplying along source contours","Multi-scale repetition that reveals new figures when viewed closer","Saturated blue, yellow and violet rolling continuously"],"works":["Exhibits and critical images that make machine misrecognition immediately visible","Music video and cover imagery that keeps a source landscape beneath a multi-scale hallucination"],"recipe":{"type":"Add type after processing, keeping it short, bold and simple enough not to fight the detail.","layout":"Preserve the source's large composition and amplify related features at several scales to create recursion into depth.","material":"Choose layers in a trained vision model and iteratively increase their activation through gradient ascent."},"avoid":"It is not a swirl filter or pasted dog faces. Preserve which layer and feature were amplified and how they bind to the source.","study":["Feature visualization and layer selection","Iterative zoom and octave scales","GoogLeNet ImageNet feature bias"]},"usage":{"license":"CC-BY-4.0","licenseName":"Creative Commons Attribution 4.0 International","licenseUrl":"https://creativecommons.org/licenses/by/4.0/","attribution":"IndexStyle — https://indexstyle.org","attributionUrl":"https://indexstyle.org","scope":"Original IndexStyle text and structured data. Reference images, quoted works, and third-party material are excluded and retain their own rights."}}