Ordered Diffusion for 3D Human Registration

2026-08-06

AI Summary

This work addresses the limitations of traditional 3D human shape registration methods, which formulate the task as a regression problem and struggle with geometric ambiguities caused by noise, occlusions, and soft-tissue deformations in scans. To overcome these challenges, the paper introduces ODin, the first approach to leverage diffusion models for this task. ODin employs an ordered diffusion mechanism to model registration uncertainty, generating point clouds that align with the target geometry while preserving the semantic structure of the template. The method conditions the generative process on a fusion of global features, local structural cues, and positional encodings to ensure semantically consistent alignment. Experiments demonstrate that ODin not only surpasses existing regression-based and highly engineered methods in accuracy—establishing a new state of the art—but also achieves a threefold improvement in registration speed.

Abstract

3D human registration has historically been treated as a regression task, assuming a unique ground-truth alignment exists between the template and an input point cloud. In reality, acquisition noise, occlusions, and unknown soft tissue dynamics introduce inherent ambiguity into human scans. Regression-based methods consequently converge to an average prediction, often failing to represent a plausible geometry. In our work, we embrace such uncertainty by modeling the registration as a distribution of alignments. We propose ODin, which formulates registration as a 3D diffusion process that generates a point cloud aligned with the target geometry while preserving template semantics through consistent point ordering. To achieve this, ODin relies on global, local, and positional conditioning, guiding each point to its correct location. Our experiments demonstrate that such a generative formulation not only outperforms its regression-based baseline, but also establishes a new state of the art, surpassing highly engineered methods while reducing the registration time by two-thirds. Pre-trained models and code are available at https://riccardomarin.github.io/odin/.

Problem

• 3D human registration
• acquisition noise
• occlusions
• soft tissue dynamics
• alignment ambiguity

Innovation

• ordered diffusion
• 3D human registration
• generative modeling
• point cloud alignment
• semantic consistency