Pith. sign in

REVIEW 2 major objections 4 minor 60 references

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

T0 review · 2 major / 4 minor · reviewed 2026-08-09 · deepseek-v4-flash

Pith's one-line read CleanPose claims that adding front-door causal adjustment and residual 3D knowledge distillation to a keypoint-based pose estimator reduces spurious correlations and yields state-of-the-art category-level pose estimation, with 61.7% on…

desk verdict CleanPose is a solid empirical recipe that overclaims its causal grounding; the front-door adjustment as implemented reduces to cross-attention. read the letter →

arxiv 2502.01312 v2 pith:4BDZKSGA submitted 2025-02-03 cs.CV

classification cs.CV
keywords category-levelobjectposeestimationcausalinferencefront-dooradjustmentknowledgedistillation3Dfoundationmodelpointcloud6-DoFspuriouscorrelation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

CleanPose argues that the remaining bottleneck in category-level object pose estimation is not feature extraction but dataset bias: training sets repeat the same objects and pose ranges, so a network learns spurious correlations that hurt novel instances. To fix this, the paper builds on the AG-Pose keypoint-based pose estimator and adds a causal module that estimates pose through front-door adjustment, using a queue of features from other training samples to block hidden confounders. It also distills category-level point-cloud knowledge from a frozen 3D foundation model (ULIP-2) through a residual projector, giving the student a debiased feature-level target. Across REAL275, CAMERA25, and HouseCat6D, the combination raises 5°2cm accuracy to 61.7%, 80.3%, and 22.4%, respectively, surpassing the prior best method by up to 4.7 points. If the causal interpretation is right, these gains come from learning the input-to-pose mechanism rather than matching the training distribution.

What carries the argument

The load-bearing object is the front-door adjustment module, a causal formula that estimates an input's effect by chaining through an intermediate mediator while bypassing hidden confounders. Given the causal graph $X \to M \to Y$ with hidden confounder $U$, the module computes $\mathbb{E}_{x'}[x'] + \mathbb{E}_{m|x}[m]$ using two attention mechanisms: self-attention on the keypoint features $m$ of the current object and cross-attention between those keypoints and a dynamic queue of $N_s$ features sampled from other training instances $x'$. The queue is updated FIFO, borrowing the dynamic queue idea of MoCo, and an adaptive weighted fusion blends the resulting feature with the original keypoint features. The second mechanism is residual knowledge distillation: the frozen ULIP-2 3D encoder (Point-BERT) provides a teacher feature, and the student's pooled point features pass through a two-MLP residual block with a zero-initialized second layer before an L2 loss aligns them, so the student is gradually pulled toward unbiased category semantics without adding confounders at initialization.

What would settle it

Train CleanPose and the AG-Pose baseline on a deliberately balanced dataset with matched instance counts and uniform pose coverage per category: if CleanPose's margin at 5°2cm mostly disappears, the gains came from correcting dataset bias, while a persistent margin would contradict that explanation. Alternatively, replace the keypoint features with random fixed features and rerun the front-door module: if accuracy stays high, the mediator assumption is false.

Watch

Extended reading notes

Core claim

The paper's central claim is that pose estimators can be made robust to confounders by replacing the observational conditional $P(Y|X)$ with the interventional $P(Y|do(X))$ via front-door adjustment, implemented with attention over a dynamic queue of cross-sample features, and by adding residual knowledge distillation from a large-scale 3D pretrained encoder. Concretely, the authors state that CleanPose achieves state-of-the-art results: 61.7% on REAL275 5°2cm (4.7% above AG-Pose), 80.3% on CAMERA25 5°2cm, and 22.4% on HouseCat6D 5°2cm, with ablations showing each component contributes. They attribute the improvements to reduced spurious correlations and more comprehensive category-level supervision for novel instances.

Load-bearing premise

The causal story depends on the keypoint features carrying all the pose-relevant information from the input, with nothing else influencing both the keypoints and the pose; the paper assumes this mediation structure without testing it.

Editorial extensions

If this is right

  • On REAL275 the method reports 61.7% at 5°2cm, 67.6% at 5°5cm, and 78.3% at 10°2cm, beating the prior-free state of the art (AG-Pose) by 4.7, 3.0, and 3.2 points respectively.
  • On CAMERA25 it reaches 80.3% at 5°2cm and 84.2% at 5°5cm, and on HouseCat6D it reaches 22.4% at 5°2cm and 24.1% at 5°5cm, a consistent gain over AG-Pose.
  • Ablations show both components matter: causal adjustment alone adds 2.7 points at 5°2cm, distillation alone adds 0.9, and together they add 4.7, with the dynamic FIFO queue outperforming a static memory bank and a similarity-based update.
  • The method keeps inference speed near the baseline (33 vs 35 frames/sec) while adding about 10% parameters, and replacing the causal module with equal-parameter MLPs reduces the gain, indicating the causal mechanism rather than extra capacity drives improvement.
  • Using a teacher encoder architecturally different from the student (Point-BERT vs PointNet++) helps, suggesting the distillation transfers category knowledge rather than shallow feature similarity.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the front-door adjustment genuinely identifies the causal effect, the same attention-over-queue module could be dropped into other geometric regression tasks that share an input-to-keypoint-to-output structure, such as shape reconstruction or grasp pose prediction.
  • The queue length ablation peaks at 80 samples and declines beyond, which suggests the module acts as a category-conditioned memory of the training distribution; a direct test is whether the gain shrinks when training data is balanced across poses and instances.
  • A stronger test of the mediation assumption would be to fix the pose head and randomize the keypoint features: if the causal module still helps, then the keypoints cannot be the sole mediator, and the gain would come from feature diversity or another mechanism.
  • The residual zero-initialized projector is a simple debiasing trick; one could test whether distilling from a teacher with broader pose coverage, rather than broader category coverage, gives larger gains on the strict 5°2cm metric.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper proposes CleanPose, a category-level object pose estimation method built on AG-Pose, with two main additions: a causal inference module based on front-door adjustment, and a residual-based knowledge distillation module that transfers features from the frozen ULIP-2 3D foundation model. The causal module represents the mediator M by learned keypoint features and approximates the front-door expectation using self-attention and cross-attention over features sampled from a dynamic queue. The distillation module aligns a residual projection of the student's pooled point features to ULIP-2 features under an L2 loss. The authors report state-of-the-art results on REAL275, CAMERA25, and HouseCat6D, with a 61.7% 5°2cm accuracy on REAL275, and provide ablations, a parameter-matched MLP control, and seed-variance experiments in the supplementary material.

Significance. If the causal interpretation were supported, the paper would make a valuable conceptual contribution by showing that front-door adjustment can mitigate dataset bias in category-level pose estimation. The empirical results are credible: the gains over AG-Pose are consistent across three benchmarks, the ablations in Tables 4 and 5 isolate the two proposed components, Table S6 includes a parameter-matched MLP replacement, and Table S5 reports low variance across sampling seeds. These elements make the empirical story substantially stronger than a bare architecture comparison. However, the central claim of the paper is the causal debiasing mechanism, and that claim is not supported by the implemented equations or by any validation of the front-door assumptions. The paper's own limitation statement in Sec. G concedes a 'gap between the causal framework and the network design,' which is precisely where the load-bearing issue lies.

major comments (2)
  1. [Sec. 4.2, Eqs. (7)-(14)] The implementation does not compute the front-door adjustment written in Eq. (7). Eq. (7) requires an outer expectation over x' under P(x'), independent of the current input x. Eq. (10) instead defines the x' weights as P(x'|g1) with g1 = q1(x), a query depending on the current input, and Eq. (12) treats this as E_{x'}[x']. The network then implements Eq. (10) as multi-head cross-attention with queries derived from Fkpt (Eqs. (13)-(14)). Unless q1 is constant, this computes E_{x'|x}[x'], not E_{x'}[x']. The module is therefore a content-addressable lookup of queue features, not the outer expectation required by front-door adjustment. The gains in Tables 4 and S6 are equally consistent with the addition of a parameterized cross-attention module; they do not validate the causal mechanism.
  2. [Sec. 3.2, Sec. 4.2, and Sec. G] The front-door assumptions are asserted, not tested. The causal graph in Fig. 2 assumes that (i) M fully mediates the effect of X on Y, and (ii) there is no unobserved confounder affecting both M and Y. Yet M is produced by a keypoint extractor trained on the same biased data, and the queue used for x' is initialized and updated with features from that same biased distribution (Sec. 4.2), so no argument establishes that its empirical distribution equals the P(x') that makes Eq. (7) identifiable. The paper's limitation statement in Sec. G explicitly acknowledges that 'the network modules in actual implementation may induce inaccuracy inevitably' and that 'such a flaw introduces a gap between the causal framework and the network design.' This concession does not repair the mismatch; it confirms that the causal interpretation is not established. The empirical contribution survives, but the central 'unbiased estimation' claim must be either substantiated with a test of the front-door assumptions or reframed as an empirical architecture improvement.
minor comments (4)
  1. [Abstract and throughout] There are repeated typographical errors: 'hightlight' in the abstract, 'confoundering effect' in Sec. 1, 'casual modeling' in Sec. 4.2, 'ontains' in Sec. 5, 'Traning Time' in Table S6, and 'debaised' in Fig. S2; these should be corrected.
  2. [Sec. 4.2, Eq. (12)] The notation in Eq. (12) is confusing: the left-hand side is written as F D(x, x') while the right-hand side uses expectations that, under the paper's own Eq. (10), depend on q1(x) and q2(x); clarifying the distinction between the ideal front-door estimator and the implemented attention-based approximation would help the reader.
  3. [Sec. 5.1, Table 1] The comparison table reports GCE-Pose from an arXiv preprint (CVPR'25) as a prior-based method; if that paper was still under review at the time of submission, it would be helpful to flag its status explicitly in the text.
  4. [Sec. 5.2 and Table S6] The parameter-matched MLP control in Table S6 is a strength, but the text does not describe the MLP architecture or its placement; a sentence explaining how the control was matched and where it was inserted would make the comparison more interpretable.

Circularity Check

1 steps flagged · score 4.0 of 10

Eq. (10) turns the front-door outer expectation over P(x') into a query-dependent softmax over P(x'|q1(x)), so the causal module is cross-attention by construction; the benchmark results themselves remain externally grounded.

  1. renaming known result [Sec. 4.2, Eqs. (7)-(14); acknowledged in Sec. G Limitation]
    "E_{x'}[x'] approximately sum over x' P(x'|g1)x' = sum_i exp(g1 x'_i^T) / sum_j exp(g1 x'_j^T) x'_i ... where g1 = q1(x) and g2 = q2(x) are two embedding functions that transmit input x into two query sets. Then, the front-door adjustment F D(.) is approximated as follows: F D(x, x') = E_{x'}[x'] + E_{m|x}[m]."

    In Eq. (7) the front-door adjustment requires averaging x' under the marginal P(x'), independent of the current input x. Eq. (10) instead weights x' by a softmax over g1 = q1(x), a query that depends on the current input, so the implemented quantity is E_{x'|x}[x'], not E_{x'}[x']. Eq. (12) then identifies this conditional attention sum with the front-door adjustment F D(x,x'), and Eqs. (13)-(14) implement it as multi-head self-attention plus cross-attention over a queue of features. The 'causal' module is therefore standard cross-attention by construction; the ablation gains in Tab. 4 and Tab. S6 show only that adding this attention module to AG-Pose helps. Section G's own limitation statement admits the gap between the causal framework and the network design.

full rationale

The benchmark results are not circular: REAL275, CAMERA25 and HouseCat6D are external test sets, the AG-Pose baseline is compared rather than being used to define the gains, and ULIP-2 is a frozen external teacher. The only concrete equation-level circularity is in the causal module. Eq. (7) is the standard front-door expression and demands E_{x'} under the marginal P(x'). Eq. (10) replaces that marginal by P(x'|g1(x)), so by construction the implemented FD(x,x') is conditional attention, and Eqs. (13)-(14) confirm it is ordinary multi-head cross-attention over a queue. Thus the paper's claim that Eq. (12) is a front-door adjustment is a renaming of attention rather than a derivation; the ablation improvements show only that this attention module helps AG-Pose, not that a causal intervention was performed. This is a genuine circular step, but it is localized: it does not infect the reported SOTA numbers, and the paper's own Section G limitation acknowledges the gap between the causal framework and the network design. Self-citations to AG-Pose and CLIPose are used as architectural baselines with independent external evaluation, so they are not load-bearing in a circular sense. Score 4.

Assumptions & free parameters 6 free parameters · 5 assumptions · 1 invented entities

The central claim rests on a small set of tunable hyperparameters and, more importantly, on unverified causal assumptions about the mediator and confounders. The hidden confounder U is a postulated construct with no independent evidence. The teacher model is external, but the definition of 'unbiased' is tied to that teacher.

free parameters (6)
  • Loss weight alpha_2 for LKD = 0.01
    Tuned via ablation in Tab. S1c; the distillation contribution depends on it.
  • Residual balancing parameter mu = 0.1
    Chosen from prior work [13]; controls the strength of the distilled residual.
  • Queue length Nq = 80
    Selected by ablation in Tab. S1a; performance peaks at 80.
  • Sampled cross features Ns = 12
    Selected by ablation in Tab. S1b; balances diversity and noise.
  • Keypoint count Nkpt = 96
    Follows the AG-Pose baseline; defines the mediator representation.
  • Pose loss weight lambda_4 = 0.3
    Taken from AG-Pose and used in the overall loss in Eq. (28).
assumptions (5)
  • domain assumption Front-door adjustment identifiability: M fully mediates X to Y and there is no unobserved confounder of M and Y.
    Required for Eqs. (5)-(7); never verified in the paper and likely violated because keypoints are learned from confounded data.
  • ad hoc to paper The causal graph X <- U -> Y and X -> M -> Y is correct for COPE, with U representing dataset bias.
    Introduced in Sec. 3.2; U is unobserved, so the graph cannot be validated from the data.
  • ad hoc to paper Linear decomposition f(x,x') = fx(x) + fx'(x') and approximation of expectations by attention weights.
    Used to reach Eq. (12); the actual network uses self/cross attention without proof that these compute the stated expectations.
  • domain assumption ULIP-2 provides unbiased category knowledge suitable as a distillation target.
    Assumed in Sec. 4.3; the teacher is itself trained on a dataset and may carry its own biases.
  • standard math Bayes' theorem and do-calculus rules.
    Used in Eqs. (1)-(7) to derive the adjustment formula.
invented entities (1)
  • Hidden confounder U
    purpose: Represents dataset bias influencing both input X and pose Y in the structural causal model.
    Introduced in Sec. 3.2; not directly measured or falsifiable from the paper's data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CleanPose: Category-Level Object Pose Estimation via Causal Learning and Knowledge Distillation." pith.science (2026). https://pith.science/paper/4BDZKSGA

@misc{pith2026250201312,
  author       = {Pith},
  title        = {Pith review of: CleanPose: Category-Level Object Pose Estimation via Causal Learning and Knowledge Distillation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4BDZKSGA}},
  note         = {Machine review of arXiv:2502.01312}
}
read the original abstract

Category-level object pose estimation aims to recover the rotation, translation and size of unseen instances within predefined categories. In this task, deep neural network-based methods have demonstrated remarkable performance. However, previous studies show they suffer from spurious correlations raised by "unclean" confounders in models, hindering their performance on novel instances with significant variations. To address this issue, we propose CleanPose, a novel approach integrating causal learning and knowledge distillation to enhance category-level pose estimation. To mitigate the negative effect of unobserved confounders, we develop a causal inference module based on front-door adjustment, which promotes unbiased estimation by reducing potential spurious correlations. Additionally, to further improve generalization ability, we devise a residual-based knowledge distillation method that has proven effective in providing comprehensive category information guidance. Extensive experiments across multiple benchmarks (REAL275, CAMERA25 and HouseCat6D) hightlight the superiority of proposed CleanPose over state-of-the-art methods. Code will be available at https://github.com/chrislin0621/CleanPose.

Figures

Figures reproduced from arXiv: 2502.01312 by the authors.

Figure 1
Figure 1. Comparison of (a) existing pose estimation approaches [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the structural causal model of COPE. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. (a) Framework of CleanPose. (b) Causal inference based on front-door adjustment is employed for mitigating potential spurious [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (2 more)
Figure 4
Figure 4. Figure 4: Comparison of ours with prior-free SOTA methods. [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Qualitative comparison on REAL275 [42]. We compare the predictions of CleanPose and the baseline AG-Pose [22]. The ground truth is marked by white borders. Performance on CAMERA25 dataset. The comparison results are presented in Tab. 2. From the observation of re￾sults…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 54 canonical work pages

  1. [1]

    Sgpa: Structure-guided prior adapta- tion for category-level 6d object pose estimation

    Kai Chen and Qi Dou. Sgpa: Structure-guided prior adapta- tion for category-level 6d object pose estimation. InProceed- ings of the IEEE/CVF International Conference on Com- puter Vision (CVPR), pages 2773–2782, 2021. 1

  2. [2]

    Fs-net: Fast shape-based network for category-level 6d object pose estimation with decou- pled rotation mechanism

    Wei Chen, Xi Jia, Hyung Jin Chang, Jinming Duan, Linlin Shen, and Ales Leonardis. Fs-net: Fast shape-based network for category-level 6d object pose estimation with decou- pled rotation mechanism. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 1581–1590, 2021. 7, 3

  3. [3]

    Secondpose: Se (3)- consistent dual-stream feature fusion for category-level pose estimation

    Yamei Chen, Yan Di, Guangyao Zhai, Fabian Manhardt, Chenyangguang Zhang, Ruida Zhang, Federico Tombari, Nassir Navab, and Benjamin Busam. Secondpose: Se (3)- consistent dual-stream feature fusion for category-level pose estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 9959–9969, 2024. 1, 2, 6, 7, 3

  4. [4]

    Gpv-pose: Category-level object pose estimation via geometry-guided point-wise voting

    Yan Di, Ruida Zhang, Zhiqiang Lou, Fabian Manhardt, Xi- angyang Ji, Nassir Navab, and Federico Tombari. Gpv-pose: Category-level object pose estimation via geometry-guided point-wise voting. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) , pages 6781–6791, 2022. 1, 2, 7, 3

  5. [5]

    ACR-Pose: Adversarial Canonical Representation Reconstruction Network for Category Level 6D Object Pose Estimation

    Zhaoxin Fan, Zhengbo Song, Jian Xu, Zhicheng Wang, Ke- jian Wu, Hongyan Liu, and Jun He. Acr-pose: Adver- sarial canonical representation reconstruction network for category level 6d object pose estimation. arXiv preprint arXiv:2111.10524, 2021. 2

  6. [6]

    Category-level 6d object pose estimation in the wild: A semi-supervised learning approach and a new dataset

    Yang Fu and Xiaolong Wang. Category-level 6d object pose estimation in the wild: A semi-supervised learning approach and a new dataset. Advances in Neural Information Process- ing Systems (NeurIPS), 35:27469–27483, 2022. 2

  7. [7]

    Open-vocabulary object detection via vision and language knowledge distillation

    Xiuye Gu, Tsung-Yi Lin, Weicheng Kuo, and Yin Cui. Open-vocabulary object detection via vision and language knowledge distillation. In International Conference on Learning Representations (ICLR), 2024. 3

  8. [8]

    Shape- constraint recurrent flow for 6d object pose estimation

    Yang Hai, Rui Song, Jiaojiao Li, and Yinlin Hu. Shape- constraint recurrent flow for 6d object pose estimation. In Proceedings of the IEEE/CVF Conference on Computer Vi- sion and Pattern Recognition (CVPR) , pages 4831–4840,

Show all 60 references
  1. [9]

    Mask r-cnn

    Kaiming He, Georgia Gkioxari, Piotr Doll ´ar, and Ross Gir- shick. Mask r-cnn. In Proceedings of the IEEE International Conference on Computer Vision (CVPR), pages 2961–2969,

  2. [10]

    Momentum contrast for unsupervised visual repre- sentation learning

    Kaiming He, Haoqi Fan, Yuxin Wu, Saining Xie, and Ross Girshick. Momentum contrast for unsupervised visual repre- sentation learning. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (CVPR) , pages 9729–9738, 2020. 2, 5

  3. [11]

    Ffb6d: A full flow bidirectional fusion network for 6d pose estimation

    Yisheng He, Haibin Huang, Haoqiang Fan, Qifeng Chen, and Jian Sun. Ffb6d: A full flow bidirectional fusion network for 6d pose estimation. InProceedings of the IEEE/CVF Confer- 4 ence on Computer Vision and Pattern Recognition (CVPR) , pages 3003–3013, 2021. 1

  4. [12]

    Distilling the knowledge in a neural net- work

    Geoffrey Hinton. Distilling the knowledge in a neural net- work. arXiv preprint arXiv:1503.02531, 2015. 3

  5. [13]

    Froster: Frozen clip is a strong teacher for open-vocabulary action recognition

    Xiaohu Huang, Hao Zhou, Kun Yao, and Kai Han. Froster: Frozen clip is a strong teacher for open-vocabulary action recognition. In International Conference on Learning Rep- resentations (ICLR), 2024. 3, 5, 6

  6. [14]

    Causalpc: Improving the robustness of point cloud classification by causal effect iden- tification

    Yuanmin Huang, Mi Zhang, Daizong Ding, Erling Jiang, Zhaoxiang Wang, and Min Yang. Causalpc: Improving the robustness of point cloud classification by causal effect iden- tification. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) , ...

  7. [15]

    Housecat6d-a large-scale multi-modal category level 6d object perception dataset with household objects in real- istic scenarios

    HyunJun Jung, Shun-Cheng Wu, Patrick Ruhkamp, Guangyao Zhai, Hannah Schieber, Giulia Rizzoli, Pengyuan Wang, Hongcheng Zhao, Lorenzo Garattoni, Sven Meier, et al. Housecat6d-a large-scale multi-modal category level 6d object perception dataset with household objects in real- i...

  8. [16]

    Adam: A method for stochastic opti- mization

    Diederik P Kingma. Adam: A method for stochastic opti- mization. arXiv preprint arXiv:1412.6980, 2014. 6

  9. [17]

    Gce-pose: Global context enhancement for category-level object pose estimation

    Weihang Li, Hongli Xu, Junwen Huang, Hyunjun Jung, Pe- ter KT Yu, Nassir Navab, and Benjamin Busam. Gce-pose: Global context enhancement for category-level object pose estimation. arXiv preprint arXiv:2502.04293, 2025. 1, 2, 7

  10. [18]

    Sar-net: Shape align- ment and recovery network for category-level 6d object pose and size estimation

    Haitao Lin, Zichang Liu, Chilam Cheang, Yanwei Fu, Guodong Guo, and Xiangyang Xue. Sar-net: Shape align- ment and recovery network for category-level 6d object pose and size estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)...

  11. [19]

    Category-level 6d object pose and size estimation using self-supervised deep prior deformation networks

    Jiehong Lin, Zewei Wei, Changxing Ding, and Kui Jia. Category-level 6d object pose and size estimation using self-supervised deep prior deformation networks. In Euro- pean Conference on Computer Vision (ECCV), pages 19–34. Springer, 2022. 1, 6, 7

  12. [20]

    Vi-net: Boosting category-level 6d object pose estimation via learn- ing decoupled rotations on the spherical representations

    Jiehong Lin, Zewei Wei, Yabin Zhang, and Kui Jia. Vi-net: Boosting category-level 6d object pose estimation via learn- ing decoupled rotations on the spherical representations. In Proceedings of the IEEE/CVF International Conference on Computer Vision (CVPR), pages 14001–14011...

  13. [21]

    Transpose: 6d object pose estimation with geometry-aware transformer

    Xiao Lin, Deming Wang, Guangliang Zhou, Chengju Liu, and Qijun Chen. Transpose: 6d object pose estimation with geometry-aware transformer. Neurocomputing, 589:127652,

  14. [22]

    Instance-adaptive and geometric-aware keypoint learning for category-level 6d object pose estimation

    Xiao Lin, Wenfei Yang, Yuan Gao, and Tianzhu Zhang. Instance-adaptive and geometric-aware keypoint learning for category-level 6d object pose estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 21040–21049, 2024. 1, 2,...

  15. [23]

    Cli- pose: Category-level object pose estimation with pre-trained vision-language knowledge

    Xiao Lin, Minghao Zhu, Ronghao Dang, Guangliang Zhou, Shaolong Shu, Feng Lin, Chengju Liu, and Qijun Chen. Cli- pose: Category-level object pose estimation with pre-trained vision-language knowledge. IEEE Transactions on Circuits and Systems for Video Technology (TCSVT), 2024....

  16. [24]

    Show, deconfound and tell: Im- age captioning with causal inference

    Bing Liu, Dong Wang, Xu Yang, Yong Zhou, Rui Yao, Zhi- wen Shao, and Jiaqi Zhao. Show, deconfound and tell: Im- age captioning with causal inference. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 18041–18050, 2022. 3

  17. [25]

    Ist-net: Prior-free category-level pose estimation with im- plicit space transformation

    Jianhui Liu, Yukang Chen, Xiaoqing Ye, and Xiaojuan Qi. Ist-net: Prior-free category-level pose estimation with im- plicit space transformation. In Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV), pages 13978–13988, 2023. 5

  18. [26]

    Mh6d: Multi-hypothesis consistency learn- ing for category-level 6-d object pose estimation

    Jian Liu, Wei Sun, Chongpei Liu, Hui Yang, Xing Zhang, and Ajmal Mian. Mh6d: Multi-hypothesis consistency learn- ing for category-level 6-d object pose estimation. IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2024. 7, 3

  19. [27]

    Structural causal 3d reconstruction

    Weiyang Liu, Zhen Liu, Liam Paull, Adrian Weller, and Bernhard Sch ¨olkopf. Structural causal 3d reconstruction. In European Conference on Computer Vision (ECCV), pages 140–159. Springer, 2022. 3

  20. [28]

    Catre: It- erative point clouds alignment for category-level object pose refinement

    Xingyu Liu, Gu Wang, Yi Li, and Xiangyang Ji. Catre: It- erative point clouds alignment for category-level object pose refinement. In European Conference on Computer Vision (ECCV), pages 499–516. Springer, 2022. 6, 7, 3

  21. [29]

    Cross-modal causal relational reasoning for event-level visual question answer- ing

    Yang Liu, Guanbin Li, and Liang Lin. Cross-modal causal relational reasoning for event-level visual question answer- ing. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), 45(10):11624–11641, 2023. 5

  22. [30]

    Re- thinking network design and local geometry in point cloud: A simple residual mlp framework

    Xu Ma, Can Qin, Haoxuan You, Haoxi Ran, and Yun Fu. Re- thinking network design and local geometry in point cloud: A simple residual mlp framework. In International Confer- ence on Learning Representations (ICLR), 2022. 8, 2

  23. [31]

    Dinov2: Learning robust visual features without supervi- sion

    Maxime Oquab, Timoth ´ee Darcet, Th ´eo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervi- sion. Transactions on Machine Learning Research J...

  24. [32]

    Causality

    J Pearl. Causality. Cambridge university press, 2009. 2, 3, 4

  25. [33]

    Causal inference in statistics: a primer

    Judea Pearl. Causal inference in statistics: a primer . John Wiley & Sons, 2016. 2, 3, 4

  26. [34]

    Sam-lad: Segment any- thing model meets zero-shot logic anomaly detection

    Yun Peng, Xiao Lin, Nachuan Ma, Jiayuan Du, Chuangwei Liu, Chengju Liu, and Qijun Chen. Sam-lad: Segment any- thing model meets zero-shot logic anomaly detection. arXiv preprint arXiv:2406.00625, 2024. 3

  27. [35]

    Pointnet++: Deep hierarchical feature learning on point sets in a metric space

    Charles Ruizhongtai Qi, Li Yi, Hao Su, and Leonidas J Guibas. Pointnet++: Deep hierarchical feature learning on point sets in a metric space. In Advances in Neural Infor- mation Processing Systems (NeurIPS), 30, 2017. 3, 6, 8, 1, 2

  28. [36]

    Learn- ing transferable visual models from natural language super- vision

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, et al. Learn- ing transferable visual models from natural language super- vision. In International Conference on Machine Learning...

  29. [37]

    Cyclical learning rates for training neural networks

    Leslie N Smith. Cyclical learning rates for training neural networks. In 2017 IEEE winter conference on applications of computer vision (WACV), pages 464–472. IEEE, 2017. 6

  30. [38]

    Shape prior deformation for categorical 6d object pose and size estima- tion

    Meng Tian, Marcelo H Ang, and Gim Hee Lee. Shape prior deformation for categorical 6d object pose and size estima- tion. In European Conference on Computer Vision (ECCV), pages 530–546. Springer, 2020. 2, 3

  31. [39]

    Least-squares estimation of transformation parameters between two point patterns

    Shinji Umeyama. Least-squares estimation of transformation parameters between two point patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , 13 (04):376–380, 1991. 2

  32. [40]

    Attention is all you need

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszko- reit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. Attention is all you need. Advances in Neural Information Processing Systems (NeurIPS), 30, 2017. 5

  33. [41]

    Densefusion: 6d object pose estimation by iterative dense fusion

    Chen Wang, Danfei Xu, Yuke Zhu, Roberto Mart ´ın-Mart´ın, Cewu Lu, Li Fei-Fei, and Silvio Savarese. Densefusion: 6d object pose estimation by iterative dense fusion. In Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 3343–3352, 2019. 1

  34. [42]

    Normalized object coordinate space for category-level 6d object pose and size estimation

    He Wang, Srinath Sridhar, Jingwei Huang, Julien Valentin, Shuran Song, and Leonidas J Guibas. Normalized object coordinate space for category-level 6d object pose and size estimation. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), ...

  35. [43]

    Vision-and-language naviga- tion via causal learning

    Liuyi Wang, Zongtao He, Ronghao Dang, Mengjiao Shen, Chengju Liu, and Qijun Chen. Vision-and-language naviga- tion via causal learning. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 13139–13150, 2024. 3, 4, 5

  36. [44]

    Phocal: A multi-modal dataset for category-level object pose estima- tion with photometrically challenging objects

    Pengyuan Wang, HyunJun Jung, Yitong Li, Siyuan Shen, Rahul Parthasarathy Srikanth, Lorenzo Garattoni, Sven Meier, Nassir Navab, and Benjamin Busam. Phocal: A multi-modal dataset for category-level object pose estima- tion with photometrically challenging objects. In Proceed- i...

  37. [45]

    Causal attention for unbiased visual recognition

    Tan Wang, Chang Zhou, Qianru Sun, and Hanwang Zhang. Causal attention for unbiased visual recognition. InProceed- ings of the IEEE/CVF International Conference on Com- puter Vision (ICCV), pages 3091–3100, 2021. 3

  38. [46]

    Unsupervised feature learning via non-parametric instance discrimination

    Zhirong Wu, Yuanjun Xiong, Stella X Yu, and Dahua Lin. Unsupervised feature learning via non-parametric instance discrimination. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR) , pages 3733–3742, 2018. 5

  39. [47]

    Ulip-2: Towards scalable multimodal pre-training for 3d understanding

    Le Xue, Ning Yu, Shu Zhang, Artemis Panagopoulou, Jun- nan Li, Roberto Mart´ın-Mart´ın, Jiajun Wu, Caiming Xiong, Ran Xu, Juan Carlos Niebles, et al. Ulip-2: Towards scalable multimodal pre-training for 3d understanding. In Proceed- ings of the IEEE/CVF Conference on Computer ...

  40. [48]

    Deconfounded image captioning: A causal retrospect

    Xu Yang, Hanwang Zhang, and Jianfei Cai. Deconfounded image captioning: A causal retrospect. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , 45 (11):12996–13010, 2021. 3, 5

  41. [49]

    Causal attention for vision-language tasks

    Xu Yang, Hanwang Zhang, Guojun Qi, and Jianfei Cai. Causal attention for vision-language tasks. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat- tern Recognition (CVPR), pages 9847–9857, 2021. 3

  42. [50]

    Point-bert: Pre-training 3d point cloud transformers with masked point modeling

    Xumin Yu, Lulu Tang, Yongming Rao, Tiejun Huang, Jie Zhou, and Jiwen Lu. Point-bert: Pre-training 3d point cloud transformers with masked point modeling. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 19313–19322, 2022. 6, 8, 1, 2

  43. [51]

    Multiple adverse weather conditions adaptation for object detection via causal intervention

    Hua Zhang, Liqiang Xiao, Xiaochun Cao, and Hassan Foroosh. Multiple adverse weather conditions adaptation for object detection via causal intervention. IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , 46 (3):1742–1756, 2022. 3

  44. [52]

    Genpose: gen- erative category-level object pose estimation via diffusion models

    Jiyao Zhang, Mingdong Wu, and Hao Dong. Genpose: gen- erative category-level object pose estimation via diffusion models. In Proceedings of the 37th International Confer- ence on Neural Information Processing Systems (NeurIPS) , pages 54627–54644, 2023. 7

  45. [53]

    What if the tv was off? examining counterfactual reasoning abilities of multi-modal language models

    Letian Zhang, Xiaotong Zhai, Zhongkai Zhao, Yongshuo Zong, Xin Wen, and Bingchen Zhao. What if the tv was off? examining counterfactual reasoning abilities of multi-modal language models. In Proceedings of the IEEE/CVF Confer- ence on Computer Vision and Pattern Recognition (C...

  46. [54]

    Omni6d: Large-vocabulary 3d object dataset for category-level 6d object pose estimation

    Mengchen Zhang, Tong Wu, Tai Wang, Tengfei Wang, Ziwei Liu, and Dahua Lin. Omni6d: Large-vocabulary 3d object dataset for category-level 6d object pose estimation. arXiv preprint arXiv:2409.18261, 2024. 2, 3

  47. [55]

    Pcp- mae: Learning to predict centers for point masked autoen- coders

    Xiangdong Zhang, Shaofeng Zhang, and Junchi Yan. Pcp- mae: Learning to predict centers for point masked autoen- coders. arXiv preprint arXiv:2408.08753, 2024. 2

  48. [56]

    Hs-pose: Hybrid scope feature extraction for category-level object pose estimation

    Linfang Zheng, Chen Wang, Yinghan Sun, Esha Dasgupta, Hua Chen, Ales Leonardis, Wei Zhang, and Hyung Jin Chang. Hs-pose: Hybrid scope feature extraction for category-level object pose estimation. arXiv preprint arXiv:2303.15743, 2023. 1, 2, 3, 7

  49. [57]

    Georef: Geometric alignment across shape variation for category-level object pose refinement

    Linfang Zheng, Tze Ho Elden Tse, Chen Wang, Yinghan Sun, Hua Chen, Ales Leonardis, Wei Zhang, and Hyung Jin Chang. Georef: Geometric alignment across shape variation for category-level object pose refinement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pat...

  50. [58]

    Semi-supervised 6d object pose estimation without using real annotations

    Guangliang Zhou, Deming Wang, Yi Yan, Huiyi Chen, and Qijun Chen. Semi-supervised 6d object pose estimation without using real annotations. IEEE Transactions on Cir- cuits and Systems for Video Technology (TCSVT) , 32(8): 5163–5174, 2021. 1

  51. [59]

    Fine-grained spatiotemporal motion alignment for contrastive video representation learning

    Minghao Zhu, Xiao Lin, Ronghao Dang, Chengju Liu, and Qijun Chen. Fine-grained spatiotemporal motion alignment for contrastive video representation learning. In Proceed- ings of the 31st ACM International Conference on Multime- dia (ACMMM), pages 4725–4736, 2023. 5

  52. [60]

    Mote: Reconciling generalization with specialization for visual-language to video knowledge transfer

    Minghao Zhu, Zhengpu Wang, Mengxian Hu, Ronghao Dang, Xiao Lin, Xun Zhou, Chengju Liu, and Qijun Chen. Mote: Reconciling generalization with specialization for visual-language to video knowledge transfer. arXiv preprint arXiv:2410.10589, 2024. 3, 5 6

Pith tools

Reviewed August 9, 2026 · model on record in the stance chip above.