PosePlaner (NeurIPS 2026): planar-geometry denoising fixes feed-forward pose predictors
zhenjun_zhao · x · 2026-10-30
Accepted at NeurIPS 2026, PosePlaner revives the classical correspondence-pose coupling via flow matching: pose refinement is cast as conditional generation conditioned on matched pairwise planar primitives and an initial pose, fusing multiple hypotheses like a continuous, data-driven RANSAC. Trained only on two-view data, it works as a periodic refiner for streaming n-view estimators without retraining. Experiments on ScanNet1500 and 7Scenes show substantial gains for SuperGlue, LoFTR, RoMA, MASt3R, PoseDiff and RayDiff, reduced long-sequence drift, and improvements over bundle adjustment in controlled comparisons.
More from Research
- Impagliazzo, Paturi and Williams launch site to scrutinize OpenAI's claimed ETH refutation — fortnow · 2026-10-30
- Know, then Draw: UCSD & Adobe Cut Image-Gen Training Compute 48x While Boosting DPG-Bench 10% — Yucheng Mao · 2026-10-30
- Live mechinterp session hunts 'neurons of sadness' in Qwen3.5 — burny_tech · 2026-10-30
- RT-SAFE Benchmark Shows Faster Reasoning Doesn't Mean Safer Agents in Real-Time Embodied Environments — Lianhuiq · 2026-10-30
- LeCun team's LeHami: symplectic Hamiltonian flow keeps JEPA world models from collapsing latent states — ylecun · 2026-10-30
- RLNF: Training AI on Human Preferences Purely From Brain Signals — alex_verem · 2026-10-30