See a live demo of isk-daemon. It allows you to browse images by visual similarity, by starting with a random selection.

No caching is used. The iskdaemon.py process takes 127mb of resident memory. Disk space used by isk-daemon database is 31mb.

Sample images used on this demo are copyrighted by Caltech 256 and includes 30,607 images, covering 256 semantic categories with at least 80 images per category.

The source code for this demo is available for browsing or for download (see the demo/ directory on the full repository). Google Web Toolkit was used on the frontend and the associated server Java component is hosted on Tomcat6. The server Java component does XML-RPC calls to the isk-daemon instance hosted on the same server. The isk-daemon instance was used to index the entire "Caltech 256" image collection.