Capturing the world for VR
- Google Photo Sphere
- Hugin
- lighweight
- easy to caputre
- realtime
- can be augmented with simple modling
- bad at expanding. Requires augmanation.
-
spawn props
-
mark floor
-
mark lines and cubes. where stuff hits the floor.
-
What can I dirive from something I know is a line?
-
Stero would proably be the best of both worlds.
-
Stero percsion with depth scales like a human (no derp)
-
or just 2d regions.
-
so draw a line on the ground
-
process it.
-
using a large number of lenses is better from a processing persepctive.
- several tools.
- faster + more relibable
- realtime
- kinect demo
- realsense demo
- slam
- manual
- very good runtime performance
- benifits from stero (Stereophotogrammetry)
- vslam
- can be realtime
- messy and unreliable
- Autodesk Memento Beta (Web-backed, Propertary)
- Autodesk 123D catch (Web-backed, Propertary, Free)
- Autodesk ReCap 360 (Web-backed, Propertary, Free for edu)
- Agisoft PhotoScan (Web-backed, Propertary, $)
- VisualSFM (Opensource, Local)
- NASA
- VSLAM (several ros packages)
- Laser range finder
- Many, many more. https://en.wikipedia.org/wiki/Comparison_of_photogrammetry_software
You need to clean up the mesh after you obtain it with photogrammetry.
- MeshLab
- Autodesk Mesh ????
- http://www.cs.jhu.edu/~misha/Code/PoissonRecon/Version5.5/
What distance should I render a photosphere?
IRL at 50ft, 2ft is impossible to tell
The resolution is 20 pixels per degree. Assume an ipd of 65mm distance is ipd
pixelTol==angleResolution*(tan((range+spacing)/ipd)-Tan(range/ipd))
In theory, I need a width of 1920*4~=8k Noyst freq? So times 2.