DICArt: Advancing Category-level Articulated Object Pose Estimation in Discrete State-Spaces
Authors: Li Zhang, Mingyu Mei, Ailing Wang, Xianhui Meng, Yan Zhong, Xinyuan Song, Liu Liu, Rujing Wang, Zaixing He, Cewu Lu
Organizations: University of Science and Technology of China · Zhejiang University · East China Normal University · Peking University · Emory University · Hefei University of Technology · Jianghuai Advance Technology Center · Anhui Provincial Key Laboratory of Humanoid Robots · Shanghai Jiao Tong University
Articulated object pose estimation is a core task in embodied AI. Existing methods typically regress poses in a continuous space, but often struggle with 1) navigating a large, complex search space and 2) failing to incorporate intrinsic kinematic constraints. In this work, we introduce DICArt (DIsCrete Diffusion for Articulation Pose Estimation), a novel framework that formulates pose estimation as a conditional discrete diffusion process. Instead of operating in a continuous domain, DICArt progressively denoises a noisy pose representation through a learned reverse diffusion procedure to recover the GT pose. To improve modeling fidelity, we propose a flexible flow decider that dynamically determines whether each token should be denoised or reset, effectively balancing the real and noise distributions during diffusion. Additionally, we incorporate a hierarchical kinematic coupling strategy, estimating the pose of each rigid part hierarchically to respect the object's kinematic structure. We validate DICArt on both synthetic and real-world datasets. Experimental results demonstrate its superior performance and robustness. By integrating discrete generative modeling with structural priors, DICArt offers a new paradigm for reliable category-level 6D pose estimation in complex environments.
Figures & tables
Figure 1: Comparison of Different Denoising Processes. We denote the rotation-related Euler angles as l,m,n , and model them using discretized bin indices for prediction. (a) illustrates the vanilla denoising process of conventional discrete diffusion models, where inconsistent convergence rates across tokens often introduce uncertainty and ambiguity in pose prediction—this can be viewed as an aggressive denoising strategy. (b) presents the reformulated denoising process proposed in this work, which is centered around a customized Flowing Mechanism . This mechanism introduces adaptive directional guidance that determines appropriate update paths for each token. It is designed to enforce consistent convergence trajectories among semantically correlated token groups, thereby enabling a more stable and gentle denoising process. Note that a rigid part ( bottle ) is chosen to illustrate the process for simplicity.
Figure 2: The Pipeline of Our DICArt. Please note that images of the object with varying saturation levels represent different degrees of noise in the pose annotations.
Category
Method
Per-part 6D Pose
Axis Descriptor
Rotation Error ( ∘ ) ↓
Translation Error (m) ↓
Angle Error ( ∘ ) ↓
Distance Error (m) ↓
Laptop
A-NCSH [ 24 ]
5.3, 5.4
0.054, 0.043
1.7
0.09
GenPose [ 48 ]
5.3, 6.1
0.068, 0.060
3.8
0.03
OP-Align [ 6 ]
6.3, 5.2
0.072, 0.058
4.5
0.11
ShapePose [ 58 ]
5.0, 4.6
0.052, 0.064
3.3
0.06
DICArt (Ours)
3.2 , 3.9
0.045 , 0.040
0.8
0.04
Table 1: Comparisons on the ArtImage Dataset. We validate our DICArt on five categories (Laptop, Eyeglasses, Dishwasher, Scissors, and Drawer) that contain 2, 3, 2, 2, 4 parts, respectively.
Figure 3: Qualitative Results on the Synthetic Dataset (left) and RGB-D Images Dataset (right).
Index
Diffusion
Rotation Error ( ∘ )
Translation Error ( m )
I
Continuous
3.1
0.143
II
Discrete
1.7
0.072
Index
Occlusion Level (Visibility)
Rotation Error ( ∘ )
Translation Error ( m )
III
0%-40%
1.8
0.089
IV
40%-80%
1.9
0.096
V
80%-100%
1.9
0.107
Table 2: Ablation Study Results . Experiments are conducted on the category Drawer .
Category
Method
Per-part Pose
Rotation Error ( ∘ )
Translation Error (m)
Box
ReArtNet [ 28 ]
3.3, 3.4
0.026, 0.031
DICArt (Ours)
1.5 , 1.6
0.009 , 0.010
Stapler
ReArtNet [ 28 ]
5.1, 6.5
0.035, 0.031
DICArt (Ours)
2.5 , 3.2
0.008 , 0.012
Cutter
ReArtNet [ 28 ]
3.3, 3.0
0.019, 0.014
Table 3: Pose Estimation Results on ReArtMix Dataset .
Per-part Rotation Error ( ∘ )
Part ID
1
2
3
4
5
6
7
A-NCSH [ 24 ]
7.8
7.9
10.3
10.5
11.2
16.4
23.5
DICArt (Ours)
1.6
4.8
7.8
7.9
8.1
12.3
15.1
Per-part Translation Error (m)
Part ID
1
2
3
4
5
6
7
A-NCSH [ 24 ]
0.012
0.044
0.067
0.066
0.179
0.236
0.403
Table 4: Pose Estimation Results on 7-part RobotArm Dataset.
Figure 4: Qualitative Results on 7-part RobotArm Dataset.
Hubei Key Laboratory of Intelligent Robot, Wuhan Institute of Technology, Wuhan, Hubei, China · University of Science and Technology of China, Hefei, Anhui, China · Peking University, Beijing, China