arXiv:2512.10674v1 Announce Type: new 
Abstract: Recent progress in zero-shot 6D object pose estimation has been driven largely by large-scale models and cloud-based inference. However, these approaches often introduce high latency, elevated energy consumption, and deployment risks related to connectivity, cost, and data governance; factors that conflict with the practical constraints of real-world robotics, where compute is limited and on-device inference is frequently required. We introduce Geo6DPose, a lightweight, fully local, and training-free pipeline for zero-shot 6D pose estimation that trades model scale for geometric reliability. Our method combines foundation model visual features with a geometric filtering strategy: Similarity maps are computed between onboarded template DINO descriptors and scene patches, and mutual correspondences are established by projecting scene patch centers to 3D and template descriptors to the object model coordinate system. Final poses are recovered via correspondence-driven RANSAC and ranked using a weighted geometric alignment metric that jointly accounts for reprojection consistency and spatial support, improving robustness to noise, clutter, and partial visibility. Geo6DPose achieves sub-second inference on a single commodity GPU while matching the average recall of significantly larger zero-shot baselines (53.7 AR, 1.08 FPS). It requires no training, fine-tuning, or network access, and remains compatible with evolving foundation backbones, advancing practical, fully local 6D perception for robotic deployment.

تم تقديم Geo6DPose كخط أنابيب خفيف الوزن ومحلي بالكامل ولا يتطلب تدريبًا لتقدير وضعية الأجسام ثلاثية الأبعاد بدون تدريب مسبق، مما يعالج قيود النماذج الكبيرة التي تعتمد غالبًا على الاستدلال السحابي، مما قد يؤدي إلى ارتفاع زمن الاستجابة واستهلاك الطاقة.

Geo6DPose se ha presentado como un pipeline ligero, completamente local y sin necesidad de entrenamiento para la estimación de pose 6D de cero disparos, abordando las limitaciones de los modelos a gran escala que a menudo dependen de la inferencia en la nube, lo que puede llevar a alta latencia y consumo de energía.

Geo6DPose a été introduit comme un pipeline léger, entièrement local et sans formation pour l'estimation de pose 6D zéro-shot, répondant aux limitations des modèles à grande échelle qui s'appuient souvent sur l'inférence basée sur le cloud, ce qui peut entraîner une latence élevée et une consommation d'énergie.

Geo6DPose has been introduced as a lightweight, fully local, and training-free pipeline for zero-shot 6D object pose estimation, addressing the limitations of large-scale models that often rely on cloud-based inference, which can lead to high latency and energy consumption.

Geo6DPose: Fast Zero-Shot 6D Object Pose Estimation via Geometry-Filtered Feature Matching

arXiv:2512.11574v1 Announce Type: new 
Abstract: Benchmarking 3D spatial understanding of foundation models is essential for real-world applications such as robotics and autonomous driving. Existing evaluations often rely on downstream finetuning with linear heads or task-specific decoders, making it difficult to isolate the intrinsic 3D reasoning ability of pretrained encoders. In this work, we introduce a novel benchmark for in-context 3D scene understanding that requires no finetuning and directly probes the quality of dense visual features. Building on the Hummingbird framework, which evaluates in-context 2D scene understanding, we extend the setup to the 3D Multi-View ImageNet (MVImgNet) dataset. Given a set of images from objects in specific angles (keys), we benchmark the performance of segmenting novel views (queries) and report the scores in 4 categories of easy, medium, hard, and extreme based on the key-query view contrast. We benchmark 8 state-of-the-art foundation models and show DINO-based encoders remain competitive across large viewpoint shifts, while 3D-aware models like VGGT require dedicated multi-view adjustments. Our code is publicly available at https://github.com/ToyeshC/open-hummingbird-3d-eval .

تم تقديم معيار جديد لتقييم الفهم المكاني ثلاثي الأبعاد لنماذج الأساس، مع التركيز على فهم المشاهد في السياق دون الحاجة إلى ضبط دقيق. يستخدم هذا المعيار مجموعة بيانات 3D Multi-View ImageNet لتقييم أداء نماذج متعددة في تقسيم وجهات نظر جديدة بناءً على مجموعة من الصور من زوايا محددة.

Se ha introducido un nuevo benchmark para evaluar la comprensión espacial 3D de los modelos de fundación, centrándose en la comprensión de escenas en contexto sin necesidad de ajuste fino. Este benchmark utiliza el conjunto de datos 3D Multi-View ImageNet para evaluar el rendimiento de varios modelos en la segmentación de nuevas vistas basadas en un conjunto de imágenes desde ángulos específicos.

Un nouveau benchmark pour évaluer la compréhension spatiale 3D des modèles de fondation a été introduit, se concentrant sur la compréhension des scènes en contexte sans nécessiter de finetuning. Ce benchmark utilise le jeu de données 3D Multi-View ImageNet pour évaluer la performance de divers modèles dans la segmentation de nouvelles vues basées sur un ensemble d'images sous des angles spécifiques.

A new benchmark for evaluating the 3D spatial understanding of foundation models has been introduced, focusing on in-context scene understanding without the need for finetuning. This benchmark utilizes the 3D Multi-View ImageNet dataset to assess the performance of various models in segmenting novel views based on a set of images from specific angles.

Geo6DPose: Fast Zero-Shot 6D Object Pose Estimation via Geometry-Filtered Feature Matching

Was this article worth reading? Share it

Artefacts.ai

4o Image Gen

Lenso.ai

GPTHumanizer

Uwear

Aqaba.ai

Ready to build your own newsroom?