The Sketchy Database: Learning to Retrieve Badly Drawn Bunnies

Re:The Sketchy Database: Learning to Retrieve Badly Drawn Bu... by DanielleMolloy on Saturday October 01, 2016 @09:24PM (#12620660)

This is a recent paper that actually creates photos from sketches of faces, using deep nets: https://arxiv.org/abs/1606.03073

Figure 4 does this for self portraits of some famous artists.

[ Parent | on HN ]

Re:The Sketchy Database: Learning to Retrieve Badly Drawn Bu... by aab0 on Saturday October 01, 2016 @09:46PM (#12620723)

You could probably apply the same code. The dataset ("acquire 75,471 sketches of 12,500 objects") sounds adequate, and if not, can be boosted by first training a CNN to do photo->sketch (throwing away information is usually easier than imagining it) and using that to boost the dataset for sketch->photo.

[ Parent | on HN ]