Did you know that you can get reasonably useable depth information from a single camera? I would have thought that it wasn’t practical. Clearly someone forgot to tell the folks at the UMichigan vision lab.
- An overview with cool video
- The first paper
- A paper from the next year, extending the concept using a Kinnect for depth
- Datasets. This kind of implies that the system is not real time?
- The code on github. No, they say it’s github, but it’s actually good, old SVN. Downloading now. Done. Big.
Looks like it should compile on any platform, and it looks like it’s not real time (images are stored in files). Looks like I need to set up a GCC environment