Accurate 6D object pose estimation is a fundamental capability for embodied agents, yet remains highly challenging in open-world environments. Many existing methods often rely on closed-set assumptions or geometry-agnostic regression schemes, limiting their generalization, stability, and real-time applicability in robotic systems. We present OMNI-PoseX, a vision foundation model that introduces a novel network architecture unifying open-vocabulary perception with an SO(3)-aware reflected flow matching pose predictor. The architecture decouples object-level understanding from geometry-consistent rotation inference, and employs a lightweight multi-modal fusion strategy that conditions rotation-sensitive geometric features on compact semantic embeddings, enabling efficient and stable 6D pose estimation. To enhance robustness and generalization, the model is trained on large-scale 6D pose datasets, leveraging broad object diversity, viewpoint variation, and scene complexity to build a scalable open-world pose backbone. Comprehensive evaluations across benchmark pose estimation, ablation studies, zero-shot generalization, and system-level robotic grasping integration demonstrate the effectiveness of OMNI-PoseX. The OMNI-PoseX achieves SOTA pose accuracy and real-time efficiency, while delivering geometrically consistent predictions that enable reliable grasping of diverse, previously unseen objects.
Figures & tables
Fig. 2: Network Architecture of the OMNI-PoseX.
Fig. 3: The figures illustrate the results of our model processing. (a) is the original RGB image. (b) is the mask segmentation result. (c) is the OMNI-PoseX prediction result.
Variant
AUC ↑
VUS ↑
Deg ↓
Sht ↓
time(ms) ∗ ↓
@25
@50
@75
5°/2cm
5°/5cm
NOCS [ 19 ]
0.0
0.0
0.0
0.0
0.0
-
-
-
SGPA [ 21 ]
10.5
2.0
0.0
4.3
6.7
-
-
-
IST-Net [ 23 ]
28.7
10.6
0.5
2.0
3.4
-
-
-
HS-Pose [ 22 ]
31.6
13.6
1.1
3.5
5.3
-
-
-
GenPose++ ∗ (w/o E) [ 25 ]
31.6
12.7
0.0
7.8
12.29
34.66
1.44
112
TABLE I: Performance evaluation on OMNI-6DPose benchmark
Category
Variant
AUC ↑
VUS ↑
@25
@50
@75
5°/2cm
5°/5cm
Rotation
Continuous representation
39.2
17.25
1.89
9.88
14.23
SO(3) representation
41.6
20.17
2.33
11.57
15.83
Fusion
DINO only
11.3
2.5
0.0
1.4
2.1
PointNet++ only
37.3
15.3
1.7
0.0
0.0
Pointwise Fusion
41.6
20.08
2.16
10.98
15.22
TABLE II: Ablation Study on OMNI-PoseX.
Fig. 4: Samples of unseen objects in Issac-Sim.
Fig. 5: Real-world demonstrations of OMNI-PoseX in daily manipulation tasks. Top : Category-level zero-shot grasping across unseen object instances, validating cross-category generalization under geometric and appearance variations. Bottom : Articulated object manipulation (drawer and cabinet door), requiring stable 6D pose tracking to support constrained motion and contact-consistent execution.
Variant
AUC ↑
VUS ↑
Deg ↓
Sht ↓
Time (ms) ↓
@25
@50
@75
5°/2cm
5°/5cm
SGPA
10.1
1.9
0.1
2.58
5.32
-
-
-
IST-Net
22.4
6.9
0.5
3.85
7.32
-
-
-
HS-Pose
25.2
7.7
0.6
4.42
9.15
-
-
-
GenPose++
33.97
11.17
1.21
6.89
15.52
38.02
3.04
178
OMNI-PoseX (Ours)
35.6
12.17
1.33
7.57
15.46
36.17
2.75
86
TABLE III: Zero-Shot Generalization on Unseen Embodied Tasks (Isaac Sim).
6D object pose estimation is fundamental to robotic manipulation and automation. Recent zero-shot methods have significantly improved generalization to unseen objects, but most still rely on large-scale pretrained models with substantial GPU computation and memory demands. These requirements complicate deployment on robotic platforms where perception, planning, and control share limited computational resources, while learned intermediate representations offer limited geometric interpretability for task-specific adaptation. To address these limitations, we propose G6D, a learning-free, geometry-driven RGB-D 6D pose solver. Given an RGB-D observation, an object instance mask, camera intrinsics, and a CAD model, G6D generates pose hypotheses through template-based geometric matching and refines them using silhouette and depth consistency, forming a purely geometry-driven pose estimation paradigm. This paradigm requires neither pretrained visual models nor target-specific training and preserves interpretable geometric representations throughout pose estimation. Moreover, adjustable hypothesis counts provide flexible accuracy-computation trade-offs, while a CPU-only configuration supports deployment without GPU resources. Experiments on LineMOD and five BOP19 datasets demonstrate advanced performance. Real-world pick-and-place experiments further demonstrate G6D's applicability to robotic manipulation. The complete project is publicly available at https://ai4control.github.io/G6D-Project-Page .
Object pose estimation is a fundamental problem in 3D vision. Although recent state-of-the-art approaches achieve strong performance, generalization to novel categories and unseen scenes remains challenging. We propose UniPose9D, a unified model for category-agnostic 9D object pose estimation: given an instance mask/ROI and either an RGB-D observation or an RGB image with predicted depth, the model estimates rotation, translation, and metric size without category labels, CAD models, mean-shape priors, or reference views. Specifically, UniPose9D samples point pairs from the observed object geometry and uses DINOv2 and PointNet features to predict NOCS coordinates for each pair. To improve accuracy, we introduce a point-pair-based RANSAC N-hop Kabsch-Umeyama algorithm with an adaptive threshold. We further employ flow matching to address symmetric ambiguities and construct a large-scale training set by curating and aligning pose annotations from existing public datasets. Experiments across eight datasets show that a single unified model achieves competitive performance on standard benchmarks while generalizing to unseen objects, unseen categories, and in-the-wild scenarios. Our code and model are available at https://github.com/qq456cvb/UniPose9D.
In many practical 6D object pose estimation scenarios, we often have access to only a single real-world RGB-D reference view per object, typically without CAD models. Existing methods largely rely on explicit 3D models or multi-view data, which limits their scalability. To address this challenging single-reference model-free setting, we propose OneViewAll, a semantic-prior-guided framework that performs pose estimation via a novel Project-and-Compare paradigm. Instead of relying on computationally expensive CAD-based rendering, our method directly aligns reference and query observations within a projection-equivariant space. OneViewAll progressively integrates hierarchical semantic priors across three levels: (1) category- and scene-level priors for efficient hypothesis initialization; (2) object-level symmetry priors for geometry completion via mirror fusion; and (3) patch-level priors for discriminative refinement. Extensive experiments demonstrate that OneViewAll achieves 92.5% ADD-0.1 accuracy on the LINEMOD dataset using only one real reference view -- significantly outperforming the CVPR 2025 baseline One2Any (52.6%). It also yields consistent improvements on YCB-V, Real275, and Toyota-Light while maintaining low inference latency. Our results underscore the efficacy of symmetry-aware projection in handling symmetric, texture-less, and occluded objects.