HiP3R: Hierarchical Pose Graphs for Long-Horizon 3D Reconstruction

Researchers present HiP3R, a hierarchical pose-graph method for long-horizon 3D reconstruction that addresses the computational and memory limits preventing feed-forward models from directly processing sequences of thousands of frames. Because long videos must be split into overlapping chunks, the division produces independent estimates that HiP3R reconciles through a hierarchical pose graph. The approach targets accurate camera motion and 3D geometry estimation across long video sequences.

Feed-forward models accurately estimate camera motion and 3D geometry from short video windows. However, computational and memory costs prevent direct inference on sequences containing thousands of frames. Long videos must therefore be processed as overlapping chunks. This division produces independ