DeepDream
In plain wordsA 2015 Google program that exaggerates whatever patterns an image-recognition network detects in a picture, producing hallucinatory images full of eyes, dog faces and swirls.
A technique that modifies an input image by gradient ascent so as to amplify the activations of chosen layers or units of a trained convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects. — originally Google's Inception classifier from the 2014 ImageNet challenge. The network's weights stay fixed while the image is repeatedly adjusted, so patterns the network faintly detects are progressively strengthened. Applied over several scales ('octaves') with smoothing priors that suppress high-frequency noise, it turns clouds and landscapes into dense, psychedelic textures of eyes, dog faces and pagodas that reflect the categories in the network's training data, much as human pareidolia sees faces in clouds. DeepDream was presented as a way to visualize what networks have learned and is closely related to feature visualizationfeature visualization, n.: Making pictures that show what a part of an image-recognition network is looking for, by repeatedly adjusting an image until that part responds as strongly as possible.; its open-source release made it one of the first neural-network art tools to reach the public.
Origin
Developed at Google by Alexander Mordvintsev, with Christopher Olah and Mike Tyka, who described it in the Google Research blog post 'Inceptionism: Going Deeper into Neural Networks' on 17 June 2015. The code was released as open source in July 2015, and 'DeepDream' became the popular name.
Related
- feature visualizationfeature visualization, n.: Making pictures that show what a part of an image-recognition network is looking for, by repeatedly adjusting an image until that part responds as strongly as possible.
- AI artAI art, n.: Images, music or other artworks made with the help of AI — from early computer drawing programs to today's prompt-based image generators — and the debate over whether and how such work counts as art.
- style transferstyle transfer, n.: Repainting a photo in the style of an artwork — your street as Van Gogh might have painted it — by keeping the photo's content and borrowing the painting's colors, strokes and textures.
- convolutional neural networkconvolutional neural network, n.: A kind of neural network built for images and other grid-like data. It slides small pattern detectors across the input, finding edges first, then shapes, then whole objects.
- hallucinationhallucination, n.: When an AI states something false or made up — a fake fact, quotation or source — fluently and confidently, as if it were true.
- interpretabilityinterpretability, n.: The effort to understand how an AI system arrives at its outputs — what information it uses and how it combines it — instead of treating it as a sealed black box.
Mentioned in
AI artAI art, n.: Images, music or other artworks made with the help of AI — from early computer drawing programs to today's prompt-based image generators — and the debate over whether and how such work counts as art., generative artgenerative art, n.: Art in which the artist invents a process — dice rolls, software, written instructions — and lets that process make the work; the phrase now also loosely covers pictures from AI image generators., style transferstyle transfer, n.: Repainting a photo in the style of an artwork — your street as Van Gogh might have painted it — by keeping the photo's content and borrowing the painting's colors, strokes and textures.