The megaworldmodel stack implements a set of tools for large-scale, database-backed object-based world mapping.

It is the successor to the worldmodel package, although it remains under active development (whereas worldmodel is basically stable). Above the capabilities provided by worldmodel, megaworldmodel is:

• Data-persistent at every step.
• Recognition-aware. Rather than work only with raw sensor data, megaworldmodel includes a generic concept of a "recognition", which includes both unsupervised "recognition" (like supporting-plane detection) and supervised recognition (including a tree-of-deformable-parts model, and TOD from the recognition_kitchen).

This package is still under heavy development, but we'd love to hear about bugs.

It's available here: https://bitbucket.org/macmason/megaworldmodel.