Skip to content

Comment on Generate images and video with Deep Dream

Comments

This reminds of the 'Magic Eye' craze, while there's an impressive technical aspect, emotionally these leave me cold, in every case I preferred the raw image to the processed one.

Yup. From a technical point of view these style transferring algorithms are really cool. But from an artistic point of view they're dull. And these examples aren't anywhere near as good as those found here: https://github.com/jcjohnson/neural-style for example. Perhaps someone can use this technique as a starting off point for creating something of artistic value, but so far I haven't seen it.

I'm quite partial to this implementation of deep dream: https://github.com/ProGamerGov/Protobuf-Dreamer

Feel like something similar should be possible with a style transfer as well.

Also: https://distill.pub/2017/feature-visualization/

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.