RIGOR Turns Panoramas Into Four-View Virtual Rigs to Fix Drift in 360° Reconstruction
zhenjun_zhao · x · 2026-09-16
Researchers from ETH Zurich (Tingjun Huang, Dmitry Rudshin, Marc Pollefeys, et al.) present RIGOR, a large-scale reconstruction pipeline for gravity-aligned omnidirectional videos (arXiv:2609.13504).
- Problem: feed-forward 3D reconstruction becomes inconsistent over long trajectories with repetitive structures, weak textures, and dynamic objects; most models don't support 360° imagery
- Method: keeps a frozen feed-forward perspective backbone and treats each panorama as a four-view virtual rig; the rig structure detects and repairs locally inconsistent predictions, retrieves loop closures via cyclic four-view consensus, and geometrically verifies revisits before global optimization
- Back-end: verified constraints drive a Sim(3) pose graph correcting accumulated rotation, translation, and scale drift
- Results: improves trajectory accuracy and geometry over a feed-forward baseline on challenging construction-site sequences
Related event: ETH Zurich's RIGOR Tackles Omnidirectional Reconstruction Drift(3 posts)→
More from Research
- Allen Institute maps topographic structure and function of locus coeruleus noradrenaline neurons in Nature — adam_k_glaser · 2026-09-17
- TMLR desk rejection rate jumps to 53%, editors quiz authors on their own papers — ducha_aiki · 2026-09-17
- Allen Institute releases 3D-MAESTRO, an automated pipeline for large-scale volumetric brain microscopy — adam_k_glaser · 2026-09-17
- Stanford's James Zou: language may be <1% of intelligence, so LLMs won't reach AGI — james_y_zou · 2026-09-17
- VC-Attention: training-free low-bit attention hits 1.9x on B200, beating FlashAttention-4 — xiuyu_l · 2026-09-17
- Boltz fused kernels deliver nearly 10x protein folding throughput gain — AllThingsApx · 2026-09-17