PoseGAM: Robust Unseen Object Pose Estimation via Geometry-Aware Multi-View Reasoning

2025-12-11

AI Summary

This work addresses the challenging problem of 6D pose estimation for unseen objects. Methodologically, we propose a geometry-aware multi-view reasoning framework that bypasses explicit feature matching: it jointly models explicit point-cloud geometry and implicit geometric representations to directly regress 6D pose from a single query image and multiple template views. To enhance cross-object generalization, we construct a large-scale synthetic dataset comprising over 190,000 diverse objects. Our key contribution is the first end-to-end integration of multi-view foundation model architecture with strong geometric priors—enabling direct, differentiable pose prediction without intermediate correspondence estimation. Extensive evaluation demonstrates state-of-the-art performance across multiple benchmarks: average AR improves by 5.1%, with up to 17.6% gain on a single dataset. The method significantly enhances robustness and generalization to previously unseen objects, particularly under domain shifts and limited training data.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

6D object pose estimation, which predicts the transformation of an object relative to the camera, remains challenging for unseen objects. Existing approaches typically rely on explicitly constructing feature correspondences between the query image and either the object model or template images. In this work, we propose PoseGAM, a geometry-aware multi-view framework that directly predicts object pose from a query image and multiple template images, eliminating the need for explicit matching. Built upon recent multi-view-based foundation model architectures, the method integrates object geometry information through two complementary mechanisms: explicit point-based geometry and learned features from geometry representation networks. In addition, we construct a large-scale synthetic dataset containing more than 190k objects under diverse environmental conditions to enhance robustness and generalization. Extensive evaluations across multiple benchmarks demonstrate our state-of-the-art performance, yielding an average AR improvement of 5.1% over prior methods and achieving up to 17.6% gains on individual datasets, indicating strong generalization to unseen objects. Project page: https://windvchen.github.io/PoseGAM/ .

Problem

• Estimates 6D pose for unseen objects robustly
• Eliminates explicit feature matching in pose estimation
• Enhances generalization using geometry-aware multi-view reasoning

Innovation

• Multi-view framework eliminates explicit feature matching
• Integrates geometry via point-based and learned representation networks
• Uses large-scale synthetic dataset for enhanced generalization