9DoF / size and pose · error

3D IoU

3D Bounding-box Intersection over Union

Compares the overlap of predicted and ground-truth 3D bounding boxes, reflecting position, rotation, and size together.

3D IoU is common in category-level 9DoF pose evaluation: when a method also predicts object size, 6DoF pose alone is insufficient.

Explore principles and formulas View related papers 37

Principle

What error does this metric penalize?

The formulas define the computation. The diagrams show whether the error concerns a 3D model, visible surfaces, the image plane, or aggregation across thresholds.

Increasing error in position, orientation, or size reduces the intersection between the two boxes.

It therefore measures overlap of occupied volumes rather than precise CAD surface-point alignment.

\operatorname{IoU}_{3D}=\frac{|B_{pred}\cap B_{gt}|}{|B_{pred}\cup B_{gt}|}

B may be an oriented 3D bounding box; implementations must specify box coordinates, dimension conventions, and how intersection volumes are computed.

Illustration created for this site to explain the metric's error space; it does not show measured data or a figure from the original paper.

How to compute

Computation steps

01
Construct the predicted 3D box from pose and size.
02
Construct the ground-truth 3D box.
03
Compute intersection and union volumes, then average over instances or apply thresholds for mAP.

Strength

What it is useful for

Includes category-level size estimation in the same score.
Easier to define than ADD when complete CAD meshes are unavailable.

Caveat

Cautions when reading results

High IoU does not guarantee accurate surface detail or critical grasping surfaces.
Different size normalization, box definitions, and IoU thresholds change the results.

Paper atlas

Which papers report this metric?

Links are derived by normalizing the metric names used in each paper. Directly reported metrics are distinguished from metrics associated through a broader protocol.

Filter the paper library

37Related papers

37Directly reported

0Linked through a protocol

2026 Directly reported

ComPose: A Unified Completion-Pose Framework for Robust Category-Level Object Pose Estimation

CVPR · Huan Ren, Yihan Chen, Chuxin Wang et al.

3D IoU

What error does this metric penalize?

Computation steps

What it is useful for

Cautions when reading results

Which papers report this metric?

ComPose: A Unified Completion-Pose Framework for Robust Category-Level Object Pose Estimation

ConceptPose: Training-Free Zero-Shot Object Pose Estimation using Concept Vectors

DynOPETs: A Versatile Benchmark for Dynamic Object Pose Estimation and Tracking in Moving Camera Scenarios

MemPose: Category-level Object Pose Estimation with Memory

Deep Learning-Based Object Pose Estimation: A Comprehensive Survey

Exploring Category-level Articulated Object Pose Tracking on SE(3) Manifolds

SE(3)-Equivariance with Geometric and Topological Guidance for Category-Level Object Pose Estimation

You Only Pose Once: A Minimalist's Detection Transformer for Monocular RGB Category-Level 9D Multi-Object Pose Estimation

Betsu-Betsu: Multi-View Separable 3D Reconstruction of Two Interacting Objects

CleanPose: Category-Level Object Pose Estimation via Causal Learning and Knowledge Distillation

GCE-Pose: Global Context Enhancement for Category-level Object Pose Estimation

Leveraging Global Stereo Consistency for Category-Level Shape and 6D Pose Estimation from Stereo Images

KeyPose: Category-Level 6D Object Pose Estimation with Self-Adaptive Keypoints

MonoDiff9D: Monocular Category-Level 9D Object Pose Estimation via Diffusion Model

R²-Art: Category-Level Articulation Pose Estimation from Single RGB Image via Cascade Render Strategy

Rethinking Correspondence-based Category-Level Object Pose Estimation

Unified Category-Level Object Detection and Pose Estimation from RGB Images using 3D Prototypes

Universal Features Guided Zero-Shot Category-Level Object Pose Estimation

You Only Estimate Once: Unified, One-Stage, Real-Time Category-Level Articulated Object 6D Pose Estimation for Robotic Grasping

Instance-Adaptive and Geometric-Aware Keypoint Learning for Category-Level 6D Object Pose Estimation

Category-level Object Detection, Pose Estimation and Reconstruction from Stereo Images

RGB-Based Category-Level Object Pose Estimation via Decoupled Metric Scale Recovery

Implicit Coarse-to-Fine 3D Perception for Category-Level Object Pose Estimation from Monocular RGB Image

GeoReF: Geometric Alignment Across Shape Variation for Category-level Object Pose Refinement

GS-Pose: Category-Level Object Pose Estimation via Geometric and Semantic Correspondence

HouseCat6D - A Large-Scale Multi-Modal Category Level 6D Object Perception Dataset with Household Objects in Realistic Scenarios

L4D-Track: Language-to-4D Modeling Towards 6-DoF Tracking and Shape Reconstruction in 3D Point Cloud Stream

LaPose: Laplacian Mixture Shape Modeling for RGB-Based Category-Level Object Pose Estimation

Living Scenes: Multi-object Relocalization and Reconstruction in Changing 3D Environments

Omni6D: Large-Vocabulary 3D Object Dataset for Category-Level 6D Object Pose Estimation

Omni6DPose: A Benchmark and Model for Universal 6D Object Pose Estimation and Tracking

OmniNOCS: A Unified NOCS Dataset and Model for 3D Lifting of 2D Objects

PACE: A Large-Scale Dataset with Pose Annotations in Cluttered Environments

SecondPose: SE(3)-Consistent Dual-Stream Feature Fusion for Category-Level Pose Estimation

U-COPE: Taking a Further Step to Universal 9D Category-level Object Pose Estimation

CenterSnap: Single-Shot Multi-Object 3D Shape Reconstruction and Categorical 6D Pose and Size Estimation

Normalized Object Coordinate Space for Category-Level 6D Object Pose and Size Estimation

Compare related metrics

n° mcm

Translation Error

Chamfer Distance