LoGeR: Long-Context Geometric Reconstruction with Hybrid Memory

TL;DR

LoGeR is a novel architecture that enables dense 3D reconstruction of extremely long video sequences by combining hybrid memory modules, significantly improving scalability and accuracy over previous methods.

Contribution

LoGeR introduces a hybrid memory system with parametric TTT and non-parametric SWA to scale dense 3D reconstruction to thousands of frames without post-optimization.

Findings

Reduces ATE on KITTI by over 74%.

Successfully reconstructs sequences up to 19,000 frames.

Outperforms prior state-of-the-art feedforward methods.

Abstract

Feedforward geometric foundation models achieve strong short-window reconstruction, yet scaling them to minutes-long videos is bottlenecked by quadratic attention complexity or limited effective memory in recurrent designs. We present LoGeR (Long-context Geometric Reconstruction), a novel architecture that scales dense 3D reconstruction to extremely long sequences without post-optimization. LoGeR processes video streams in chunks, leveraging strong bidirectional priors for high-fidelity intra-chunk reasoning. To manage the critical challenge of coherence across chunk boundaries, we propose a learning-based hybrid memory module. This dual-component system combines a parametric Test-Time Training (TTT) memory to anchor the global coordinate frame and prevent scale drift, alongside a non-parametric Sliding Window Attention (SWA) mechanism to preserve uncompressed context for high-precision adjacent alignment. Remarkably, this memory architecture enables LoGeR to be trained on sequences of 128 frames, and generalize up to thousands of frames during inference. Evaluated across standard benchmarks and a newly repurposed VBR dataset with sequences of up to 19k frames, LoGeR substantially outperforms prior state-of-the-art feedforward methods--reducing ATE on KITTI by over 74%--and achieves robust, globally consistent reconstruction over unprecedented horizons.