IUP-Pose: Decoupled Iterative Uncertainty Propagation for Real-time Relative Pose Regression via Implicit Dense Alignment v1

TL;DR

IUP-Pose introduces a decoupled, iterative, and geometry-driven framework for real-time relative pose estimation that effectively aligns cross-view features and refines rotation and translation separately, achieving high accuracy and efficiency.

Contribution

The paper presents IUP-Pose, a novel decoupled iterative approach with implicit dense alignment, enabling real-time, end-to-end trainable relative pose regression with improved accuracy and efficiency.

Findings

Achieves 73.3% AUC@20deg on MegaDepth1500.

Runs at 70 FPS with 37M parameters.

End-to-end differentiable with implicit dense alignment.

Abstract

Relative pose estimation is fundamental for SLAM, visual localization, and 3D reconstruction. Existing Relative Pose Regression (RPR) methods face a key trade-off: feature-matching pipelines achieve high accuracy but block gradient flow via non-differentiable RANSAC, while ViT-based regressors are end-to-end trainable but prohibitively expensive for real-time deployment. We identify the core bottlenecks as the coupling between rotation and translation estimation and insufficient cross-view feature alignment. We propose IUP-Pose, a geometry-driven decoupled iterative framework with implicit dense alignment. A lightweight Multi-Head Bi-Cross Attention (MHBC) module aligns cross-view features without explicit matching supervision. The aligned features are processed by a decoupled rotation-translation pipeline: two shared-parameter rotation stages iteratively refine rotation with uncertainty, and feature maps are realigned via rotational homography H_inf before translation prediction. IUP-Pose achieves 73.3% AUC@20deg on MegaDepth1500 with full end-to-end differentiability, 70 FPS throughput, and only 37M parameters, demonstrating a favorable accuracy-efficiency trade-off for real-time edge deployment.