This is an overview of some of the techniques I’ve been playing around with this week. As well as an look at how I went from a sketch, to maquette, to a 3d proxy model. One of my goals with all of this experimentation was to find a more automated approach to reconstructing a model from photos or video. The last time I had experimented with this process was years ago with Autodesk’s Image Modeler(not fun), things have come a long way since then.
This character is based on the Japanese Spider Crab. In reality they can grow to be over 2 meters long, but this crab character will be much bigger, the size of a house! In “Sayonara” (tentative title) he represents, change, chaos, and things we can’t control in life. In the animation he will be a mix of an actual puppet (this maquette) and cg augmentations such as legs, a mouth and other creepy things that wiggle out.
For the 3d reconstruction I tried several applications, only 3 really worked for this purpose;
(pretty amazing $20 program)
-poor concave surface reconstruction
-registration matte is limiting
-only able to scan one side at a time
-completely automated registration
-need only to upload your photos
-wasn’t able to reconstruct some parts
-point cloud detail was low
(the clear winner, oh and free)
-upload photos/completely automated
-accurate/dense point cloud
-even reconstructed convex areas quite well
-no export/”hacky” to get data
In the end I decided to use the data from Photosynth. Although exporting the point cloud was a bit technical. There are a couple of ways to hack the data out of the Microsoft servers(hopefully an export feature will be added soon);
http://binarymillenium.com/2008/08/photosynth-export-process-tutorial.html(I used this process.)
Lastly there’s a handy free tool called meshlab, that does a great job of rebuilding a surface from the point cloud. I was able to export an obj and send it right to Maya. Here’s a guide for meshing your point cloud in meshlab;
Ok, time to get back to work. Comments and questions are much obliged!