The video pulls off several tricks that are each, by their own, pretty cool:
- Morphable model of the human face that you adjust sliders to deform (e.g. male/female, weight)
- Overlaying their morphable model on a photograph (e.g. Tom Hanks) and automatically extracting a 3D model.
- Manipulating their 3D model inside the photograph to change posing, facial expression, facial features, etc... as well as adding 3D objects (e.g. hat) to the photo and using the 3D model to correctly determine lighting.
via Table of Malcontents/work e-mail list
Comments (1)
I saw a presentation over the summer (at the BA Festival of Science in York) of similar technology being used to accurately create three dimensional models of the faces of children with various genetic disorders.
By averaging large numbers of diagnosed children with the same disability or genetic disorder, you can create an averaged 'face' of that disorder against which other children can be measured. Trends quickly show up, including 'twisting' effects in the skulls of children with certain disorders, where one side of the skull seemes to be pushed forwards and the other retreating. Similarly natural expressions (derived from mapped models) can be averaged for such disorders, creating first impression diagnosis tools for rarer disorders which new doctors and psychologists may not otherwise have experience with.
Posted by Paul Schofield | November 5, 2007 9:07 AM
Posted on November 5, 2007 09:07