A great paper (Weta digital were involved) on moving facial performances onto many types of models using blendshapes and spacetime to perform a smooth mapping to totally different geometry. Very cool.
A cool paper on compositing images into a scene with realistic lighting. Quite impressive how much they were able to do with a source and destination 2D image.
This paper was pretty mindblowing and used a technique described as melding to sample other parts of the image in many ways (colour, scale, rotation, gain, bias…) then be able to merge or infill. Check out how they replaced the magenta pixels in this ceiling… AUTOMATICALLY!!.
Phasespace had their new high res, high speed cameras on display. They were using a LED lighting rig so they could shorten the exposure time indoors. It’s still much faster in daylight, they had samples freezing the seam on a baseball at 200 frames per second!
This is a mosaic of the six cameras trained on a guitar player. You could point them all at a face and then use it to reconstruct a 3D model.. so many possible uses.
Then this is a zoom into the detail in one of the images.. all captured at high speed. This detail, hairs, the pores of the skin is what makes 3D correspondance possible for great reconstructions.
As I did last year, I can highly recommend the advanced real time rendering course [notes here] which was a great selection of innovative teams talking about what they were doing in detail in regard to multi cores, gpus, scene management, lighting and rendering.