1QueryimagesResultimagesAutomatic generalizationDataset: wikiart images - https://www.wikiart.org/.No image pre-processing performed. No textual data used.Feature vectors generated automatically with deep learning.www.xyggy.com © 2019
View Slide
2Walkthrough …www.xyggy.com © 2019Query image is womanwith long neck, smalleyes and narrow faceleaning to one side.Results (left to right) arein character with queryimage.Note: most art is uniqueand so identicalsimilarity will be rare.
3www.xyggy.com © 2019Add 2nd query image.Given one or moreimages, finds otherrelevant images inranked order.Industry focus is onsqueezing every pixelfrom image butmetaphorically twoeyes are better thanone.
4www.xyggy.com © 2019A modern art couple is3rd query image.First result is duplicateof modern couple. The2nd is in style of thelong-necked woman.Remaining results are instyle of modern couple.Portraits #3 and #8 arein character with firsttwo query images.
5www.xyggy.com © 2019The 4th query image iswoman with long’ishneck and headscarf.Results include femaleswith head-coverings,long necks and leaningheads.Result #8, in style ofmodern couple, haslong neck and face.
6In a nutshell …www.xyggy.com © 2019Given one or more items,Thingy finds other relevantitems in ranked order
7What is a thing?www.xyggy.com © 2019Items (“things”) can be images, audio,video, db records, documents, genes …in fact, all datatypes.Multi-modal? Yes, eventually.
8Dynamic AIwww.xyggy.com © 20191. Automatic generalization – in realtime2. Interactive – query with one or more items;add and remove items from query. Givesconsumers control3. Dynamic – add new data at anytimeAnd, much more …
9www.xyggy.com © 2019Thingy