Strength
What it is useful for
- Directly related to visual appearance in the image.
- Requires no depth map and supports RGB-only protocols.
6DoF / image projection · error
2D Projection
2D Projection Error
Projects CAD points into the image and measures the average pixel distance between predicted and ground-truth projections.
Measures visible misalignment from the camera's perspective, often as success rate at < 5px; useful for RGB-only methods and AR visual alignment.
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.
A small 3D change can produce a large pixel displacement near the camera or with a long focal length.
It therefore indicates whether the final rendering, overlay, or image alignment is visibly displaced.
How to compute
Computation steps
- 01
Transform the CAD points using both poses.
- 02
Project both 3D point sets with the same camera intrinsics.
- 03
Average pixel distances, or compute the proportion below 5px.
Caveat
Cautions when reading results
- A 5px threshold has different strictness in 640px and 4K images.
- Pixel error is not equivalent to 3D grasping or assembly error.
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.
17Related papers
17Directly reported
0Linked through a protocol
2026 Directly reported
Deep Learning-Based Object Pose Estimation: A Comprehensive Survey
IJCV · Jian Liu, Wei Sun, Hui Yang et al.
Original terminology: 2D Projection
2025 Directly reported
BoxDreamer: Dreaming Box Corners for Generalizable Object Pose Estimation
ICCV · Yuanhong Yu, Xingyi He, Chen Zhao et al.
Original terminology: Proj2D@5px
2025 Directly reported
Environment-Agnostic Pose: Generating Environment-independent Object Representations for 6D Pose Estimation
ICCV · Shaobo Zhang, Yuhang Huang, Wanqing Zhao et al.
Original terminology: 2D Projection
2025 Directly reported
GS-Pose: Generalizable Segmentation-Based 6D Object Pose Estimation With 3D Gaussian Splatting
3DV · Dingding Cai, Janne Heikkilä, Esa Rahtu
Original terminology: 2D Projection
2025 Directly reported
iG-6DoF: Model-free 6DoF Pose Estimation for Unseen Object via Iterative 3D Gaussian Splatting
CVPR · Tuo Cao, Fei Luo, Jiongming Qin et al.
Original terminology: Projection 2D
2025 Directly reported
ONDA-Pose: Occlusion-Aware Neural Domain Adaptation for Self-Supervised 6D Object Pose Estimation
CVPR · Tao Tan, Qiulei Dong
Original terminology: 2D projection error
2024 Directly reported
MFOS: Model-Free & One-Shot Object Pose Estimation
AAAI · JongMin Lee, Yohann Cabon, Romain Brégier et al.
Original terminology: 2D Projection / Proj2D (5px)
2022 Directly reported
Gen6D: Generalizable Model-Free 6-DoF Object Pose Estimation from RGB Images
ECCV · Yuan Liu, Yilin Wen, Sida Peng et al.
Original terminology: Prj-5
2022 Directly reported
MegaPose: 6D Pose Estimation of Novel Objects via Render & Compare
CoRL · Yann Labbé, Lucas Manuelli, Arsalan Mousavian et al.
Original terminology: 2D Projection / Proj2D@5px
2022 Directly reported
PIZZA: A Powerful Image-only Zero-Shot Zero-CAD Approach to 6 DoF Tracking
3DV · Van Nguyen Nguyen, Yuming Du, Yang Xiao et al.
Original terminology: 2D Projection / Proj2D-5px
2022 Directly reported
YOLOPose: Transformer-based Multi-Object 6D Pose Estimation using Keypoint Regression
arXiv · Arash Amini, Arul Selvam Periyasamy, Sven Behnke
Original terminology: 2D Projection
2019 Directly reported
DPOD: 6D Pose Object Detector and Refiner
ICCV · Sergey Zakharov, Ivan Shugurov, Slobodan Ilic
Original terminology: 2D Projection
2019 Directly reported
PVNet: Pixel-wise Voting Network for 6DoF Pose Estimation
CVPR · Sida Peng, Yuan Liu, Qixing Huang et al.
Original terminology: 2D Projection
2018 Directly reported
Deep Object Pose Estimation for Semantic Robotic Grasping of Household Objects
CoRL · Jonathan Tremblay, Thang To, Balakumar Sundaralingam et al.
Original terminology: 2D Projection
2018 Directly reported
PoseCNN: A Convolutional Neural Network for 6D Object Pose Estimation in Cluttered Scenes
RSS · Yu Xiang, Tanner Schmidt, Venkatraman Narayanan et al.
Original terminology: 2D Projection
2018 Directly reported
Real-Time Seamless Single Shot 6D Object Pose Prediction
CVPR · Bugra Tekin, Sudipta N. Sinha, Pascal Fua
Original terminology: 2D Projection
2017 Directly reported
SSD-6D: Making RGB-based 3D Detection and 6D Pose Estimation Great Again
ICCV · Wadim Kehl, Fabian Manhardt, Federico Tombari et al.
Original terminology: 2D Projection
Next
Compare related metrics
Compare metrics at the same level to understand why one score improves while another stays unchanged.
ADD
Average Distance of Model Points
error
Compares each CAD point under predicted and ground-truth poses, averaging the 3D distances.
MSPD
Maximum Symmetry-Aware Projection Distance
error
Measures the worst model-point displacement in image pixels after accounting for symmetry.
n° mcm
Rotation–Translation Threshold Accuracy
threshold
Counts a pose as correct only when both rotation and translation errors fall below their thresholds.