COMPOSE: Hypergraph Cover Optimization for Multi-view 3D Human Pose Estimation

2026-06-08 · Source: cs.CV updates on arXiv.org · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Expert, extended

Summary

COMPOSE is a new framework for multi-view 3D human pose estimation, addressing the limitations of traditional pairwise association methods. It reformulates the correspondence matching as a hypergraph partitioning problem, enabling simultaneous modeling of higher-order relationships across multiple camera views. This approach enhances robustness against noisy 2D detections and occlusions. While the underlying integer linear program has theoretical exponential complexity, COMPOSE employs an efficient geometric pruning strategy to reduce the search space, making it practically solvable. The framework demonstrates substantial performance gains, achieving improvements of up to 23% in average precision over previous optimization-based methods and up to 11% over self-supervised end-to-end learned methods. Evaluated on datasets like CMU Panoptic, Shelf, and Campus, COMPOSE shows superior accuracy, including an AP25 increase from 37.63 to 54.66 and an MPJPE of 23.62mm on CMU Panoptic, with an average runtime of 22.7ms per frame.

Key takeaway

For Computer Vision Engineers developing multi-view 3D human pose estimation systems, you should consider hypergraph-based approaches like COMPOSE. This method offers superior accuracy and robustness in complex, occluded scenes, surpassing traditional pairwise association or self-supervised learning. It reduces reliance on extensive 3D annotated datasets. Implement this framework to achieve more reliable 3D pose reconstructions for applications like human-robot interaction or sports analysis.

Key insights

Hypergraph partitioning for multi-view 3D pose estimation improves accuracy and robustness over pairwise methods.

Principles

• Global consistency resolves local ambiguities.
• Higher-order relationships enhance robustness.
• Geometric pruning can tame exponential complexity.

Method

COMPOSE detects 2D keypoints, constructs a weighted hypergraph, solves a weighted exact cover problem via ILP with geometric pruning, then triangulates 3D poses.

In practice

• Apply hypergraph models for multi-sensor fusion.
• Use geometric pruning to optimize NP-hard problems.
• Improve 3D pose accuracy in occluded scenes.

Topics

• 3D Human Pose Estimation
• Multi-view Systems
• Hypergraph Optimization
• Integer Linear Programming
• Geometric Pruning
• Pose Correspondence

Best for: Research Scientist, AI Scientist, Computer Vision Engineer, Machine Learning Engineer

Related on AIssential

• DisPOSE: Projected Polystochastic Diffusion for Self-Supervised Multi-View 3D Human Pose Estimation — Self-supervised multi-view 3D human pose estimation can be achieved by modeling discrete person assignment as projected diffusion on polystochastic tensors.
• Mesh-Aware Epipolar Matching for Multi-View Multi-Person 3D Pose Estimation in Basketball — Dense mesh geometry significantly enhances training-free multi-view 3D pose estimation in complex sports environments.
• Geometric-Aware Hypergraph Reasoning for Novel Class Discovery in Point Cloud Segmentation — Hypergraph reasoning and geometric-aware prototypes improve novel class discovery in point cloud segmentation by capturing high-order and spatial relationships.
• SToRe3D: Sparse Token Relevance in ViTs for Efficient Multi-View 3D Object Detection — Jointly sparsifying 2D image tokens and 3D object queries with relevance-aligned filtering significantly boosts multi-view 3D detection efficiency.
• View-Consistent 3D Scene Editing via Dual-Path Structural Correspondense and Semantic Continuity — Achieving view-consistent 3D scene editing requires joint distribution modeling across viewpoints.
• Every9D-21M: Large-Scale Real-World 9D Canonicalization of Everyday Objects — Large-scale, real-world 9D pose datasets can be built efficiently by propagating sparse manual annotations across object-centric videos.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.