| This looks too good to be true. - My first question is if this is machine learbed or algorithm. I doubt its algorithm as it needs "object detection" then needs to know the cloths "resting state". - Does it detect changes based on shadows or are the multiple cameras? If shadows, this may be gameable/corruptable. If multiple cameras, how many? - I saw xy rotation, xy skewing and multiple objects. Folding in the xz plane and could detect a "front and back" to a t-shirt when he did the jump spin. I didnt see how it adapted to xz rotation or a combination. I also dont understand how it knows there was a "back" and a "front" without mapping the object in internal 3d spacewhich may or may not work for some arbitrary shape. I did not see adaption to depth though in great degreee If I understand correctly - machine learned arbitrary 3d blobs into 3d space. Through multiple cameras or lighting or blth - machine learn changes to the blob as not "new blobs" but changes to existing ones - create image to match that of location and polygon of the fabric. Merge certain pixels when condensing. Duplicating when expanding. Employ anti aliasing techniques - send image to projector Im not sure the application but it is certain unique, fun and interesting technology. |