Show HN: Mediachain Attribution Engine | Hacker News Reader

Show HN: Mediachain Attribution Engine

images.mediachain.io

64 pointsdenisnazarov8 comments

Hey HN!

I'm Denis, one of the people working on Mediachain! We're really excited to launch Mediachain Attribution Engine today!

We think it's the best image search engine for creators. You can find free, high quality images that you can share and re-use, with attribution automatically baked in.

Mediachain Attribution Engine makes it easy to find a great image that you can feel good about sharing or using in your blog post, presentation, website, etc.

You can upload any image from the web to find out who made it and where it came from. If Attribution Engine doesn’t know the creator, it’ll suggest visually similar images that are licensed for re-use and give credit to the author.

Attribution Engine is the first application powered by decentralized, open data in Mediachain and is built on top of the newly launched protocol architecture (v1.0).

Creators can sign up to register works and developers can use the data in their own apps, or contribute directly to Attribution Engine by following our quick start guide.

Learn more on our blog!

https://blog.mediachain.io/introducing-mediachain-attributio...

https://blog.mediachain.io/mediachain-v1-0-be2b8fa2153

I noticed that the images appear to be cropped to a specific aspect ratio from the originals. Is there a way to embed the originals, or is it done the way it is for simplicity?

Good catch, this was done as an optimization for a few sources. We are working to backfill the originals in the next few days, as our goal is to preserve the original images in their pristine from. Should not affect the majority of the collection. For now you should be able to click to the source and find the original image there. Sorry for the inconvenience!

For how Attribution Engine uses neural networks to learn image aesthetics & give creators credit, checkout the open source repo on Github: https://github.com/mediachain/mediachain-indexer

Interesting.

How do you guard against people fraudulently claiming they own images?

Because our system represents extant authorship relationships (as opposed to creating a whole new first-to-file system like some other approaches in the space), we defer to existing authorities whenever possible -- so authorship information provided by MoMA, NYPL, DPLA, etc is considered authoritative, whereas statements by a private individual generally do not carry the same weight. All of this (including conflicting statements) is surfaced at read time and can be subject to different weighting depending on the situation (best-effort attempt to find the author of an image for discovery purposes, versus automatic payment, for example)

I love the idea of this!

Excellent!

Reply on news.ycombinator.com