search, dedupe, and media ingestion for mediachain
Python
Latest commit da1e482 Oct 27, 2016 @autoencoder autoencoder Readme text.
Permalink
Failed to load latest commit information.
images_gh
mediachain Readme. Oct 27, 2016
.gitignore Model and task queue updates. Oct 24, 2016
INSTALL.md Readme and Install. Oct 24, 2016
MANIFEST.in Model and task queue updates. Oct 24, 2016
README.md
apidoc.json
requirements.txt Model and task queue updates. Oct 24, 2016
setup.py Model and task queue updates. Oct 24, 2016

README.md

Search & Attribution for Mediachain

Mediachain Indexer Core is a state of the art, high-relevance, aesthetics-weighted, image search engine and reverse image search for the world's 400 million+ Creative Commons licensed images.

Inside the models

Mediachain neural aesthetics model predictions for high (top) and low (bottom) image aesthetics:

Achieving a human-like sense of aesthetics, while balancing with the need for high-relevance results is a significant challenge. The Mediachain Indexer achieves this using the latest in end-to-end supervised deep neural networks.

Better images from larger collections

Sample calculations of the aesthetics model:

This combination of models gives the Mediachain Indexer results quality similar in aesthetics to the much smaller hand curated collections, while taking advantage of far larger un-curated Creative Commons resources.