We have generated data sets for anyone wanting to continue research. The synthetic sets were generated largely with the scene and image classes, found on the code page. All of the synthetic data includes objects that are horizontally convex. They also only move horizontally and all have ten frames. The data folder is also accessible.

The photographic sets are far more difficult. Included is three version of each frame: the original picture (renamed and converted to .ppm), the output of the segementer using good values, and the output of the photographic rasterizor. The last is easiest to use, but better algorithms could probably get better data from the original photographs. Special thanks to Gonzalo Alonso for help taking the stereo setups.

Synthetic Data Sets:
All sets

One Dimensional Sets:
Mondrian Motion, integers velocities only: A | B | C
Mondrian Motion, all velocities: A | B | C

Two Dimensional Sets:
Mondrian Motion, all velocities: A | B | C
Mondrian Stereo, all velocities: A | B | C

Photographic Data Sets:

Adirondacks
Nine frames of Mondrian Stereo, largely straight lines.
Horizontally convex. The easiest convex set.

Rockies
Nine frames of Mondrain stereo. Mostly horizontally convex.

Alps | Himalayas
Five frames each. Very similar. Non-convex shapes.

Special thanks to Gonzalo Alonso '06 for his help taken data sets. The sets were taken with a Canon digital Rebel using the setup for MiddROVR's Stereo Vision research.