Pith. sign in

REVIEW 2 major objections 5 minor 38 references

A medium-scale set of human and animal 3D shapes plus wild images is enough for unsupervised pixel-wise left-right labeling that works even on cars and trains.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.5

2026-07-11 10:12 UTC pith:NAGAEBTJ

load-bearing objection First unsupervised dense pixel-level left-right maps for wild images; hybrid 3D+image training works and generalizes past the training categories. the 2 major comments →

arxiv 2607.05006 v1 pith:NAGAEBTJ submitted 2026-07-06 cs.CV

Unsupervised Pixel-Level Semantic Left-Right Understanding of In-the-Wild Images

Weikang Wang , Tobias Wei{\ss}berg , Florian Bernard This is my paper
classification cs.CV
keywords semantic left-rightpixel-level predictionunsupervised learninghybrid 3D-image trainingvision foundation modelsreflective symmetryin-the-wild images
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Pixel-level left-right labeling of ordinary photographs is hard because a single image lacks 3D geometry and is full of occlusions, partial views and pose changes. This paper shows that the difficulty can be removed without any left-right annotations. The method renders a modest collection of human- and quadruped-animal meshes, extracts features from a frozen vision foundation model on both those renderings and ordinary photographs, and trains a lightweight predictor with purely geometric consistency losses on the meshes together with a simple left-right flip-consistency loss on the photographs. The resulting network produces dense left-versus-right maps that stay coherent across object categories never seen in the 3D data (cars, trains, aeroplanes) and across painting or anime styles. The practical claim is that left-right semantics already latent in modern vision models become reliable once a small amount of 3D geometric prior is supplied, opening a route to unsupervised chirality understanding for any image.

Core claim

A hybrid unsupervised training scheme that jointly uses a medium-scale 3D shape collection of mainly human- and animal-like objects and diverse in-the-wild images is sufficient to train a pixel-wise predictor that produces accurate semantic left-right maps for single-view photographs, including categories and styles never present in the 3D data.

What carries the argument

Pix2LR: a light-weight pixel predictor trained by swapping the order of left-right prediction and surface aggregation so that the same network can be supervised by 3D geometric losses (dissimilarity, total variation, fifty-fifty) on rendered shapes and by an image-based flip-consistency loss on masked wild photographs.

Load-bearing premise

Vision foundation models already contain sufficiently rich and category-consistent left-right semantics that a modest set of human and animal shapes can transfer those semantics to arbitrary man-made objects and image styles.

What would settle it

Train the identical hybrid pipeline on the same wild-image sets but replace the human/animal shape collection with an equal number of meshes drawn only from rotationally symmetric or left-right-ambiguous categories (bottles, tables); if dense left-right accuracy on cars, trains and aeroplanes collapses while remaining high on humans, the transfer hypothesis is falsified.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Dense left-right maps can be used as free supervision to resolve left-right ambiguities in monocular pose estimation and keypoint detectors.
  • Image generation and editing models can be regularised by an explicit left-right consistency term derived from the same predictor.
  • Category-level part segmentation becomes possible without left-right labels once the predictor supplies the missing chirality signal.
  • The same hybrid recipe can be applied to other latent geometric attributes (front-back, up-down) that foundation models may already encode.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If the transfer works for left-right, a similar surface-aggregation plus flip-consistency loop may extract other extrinsic symmetries without 3D annotations for those symmetries.
  • The method implicitly treats the camera-facing side as the semantic front; this convention could be turned into an explicit front-back predictor with only minor architectural change.
  • Failure modes on thin structures (bicycle handlebars) suggest that adding a light attention block over instance masks would be a low-cost next experiment.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper introduces Pix2LR, an unsupervised framework for dense pixel-wise semantic left-right labeling of objects in single-view in-the-wild images. It jointly trains on a medium-scale 3D shape collection (BeCoS: mainly human- and quadruped-animal meshes) and multi-category image sets (PF-Pascal, SPair-71k). Rendered multi-view images and wild images are processed by frozen vision foundation models (Stable Diffusion + DINOv3); a lightweight predictor maps features to per-pixel left-right scores in [-1,1]. For shapes, predictions are aggregated onto vertices so that the geometric losses of χ (dissimilarity, total variation, fifty-fifty) can be applied; an additional image-level consistency loss enforces agreement between original and horizontally flipped wild images. The resulting model produces dense left-right maps that generalize to categories never seen in the 3D prior (cars, trains, aeroplanes) and to non-photographic styles, outperforming SD+DINO, GeoAware and a re-purposed χ baseline on both keypoint-wise wild-image metrics and dense rendered-image metrics.

Significance. Pixel-level semantic left-right understanding of unconstrained images has been under-explored; existing 2D methods either detect only extrinsic axes or refine sparse correspondences under keypoint supervision. Pix2LR supplies the first unsupervised dense solution and demonstrates that a modest, category-limited 3D prior plus foundation-model features is already sufficient for strong cross-category transfer. The empirical gains on SPair-71k, PF-Pascal and BeCoS (Tables 2–4), the ablations confirming each loss and the multi-category 3D prior (Table 5), and the zero-shot qualitative results on paintings, anime and unseen classes (Fig. 6) constitute concrete, reproducible evidence. The work therefore opens a practical route to left-right-aware features for pose estimation, correspondence and generative editing without requiring dense 2D annotations.

major comments (2)
  1. The evaluation protocol for wild images (Sec. 4, Eq. 7) relies on newly introduced keypoint-wise left-right labels that the authors themselves assigned by manually partitioning the existing sparse keypoints of each category into left/right/boundary. While the procedure is transparent, no inter-annotator agreement or sensitivity analysis is reported. Because these labels form the sole quantitative ground truth for Tables 2 and 3, a short reliability check (e.g., second annotator on a subset, or leave-one-keypoint-out stability) would strengthen confidence that the large reported margins are not partly an artifact of the labeling convention.
  2. The hybrid-training hypothesis (Sec. 3.2) asserts that human/quadruped geometry supplies a sufficient 3D prior for arbitrary object categories. Table 5 shows that replacing BeCoS by FAUST measurably hurts wild-image accuracy, yet the paper never tests a still more restricted prior (e.g., only humans) or a deliberately mismatched prior (e.g., only chairs). A single additional ablation of this form would make the claimed sufficiency of the medium-scale animal/human prior more falsifiable and would clarify how much category diversity is truly required.
minor comments (5)
  1. Notation table (Table 1) lists χ_M, χ_v etc. but the main text sometimes uses χ without the subscript; a uniform convention would improve readability.
  2. Fig. 3(b) caption is dense; a short sentence clarifying that the order swap is what enables the wild-image path would help readers who skip the surrounding text.
  3. The loss weights λ1–λ4 and the number of rendered views Nr are free parameters; their concrete values (or a statement that they were fixed once and never tuned per dataset) should appear in the main text or a short implementation paragraph, not only in the supplementary.
  4. In Sec. 5 the authors note that thin structures (bicycle handlebars) remain difficult; a quantitative breakdown of accuracy on such thin parts, if available, would be a useful addition to the limitations discussion.
  5. A few typographical inconsistencies appear (e.g., “PixLR” vs. “Pix2LR” in Sec. 4.3; missing spaces after some citations). A careful proof-reading pass is recommended.

Circularity Check

1 steps flagged

No significant circularity: unsupervised consistency losses plus independent evaluation; only minor methodological self-citation of prior χ losses.

specific steps
  1. self citation load bearing [Section 3.1 Background; Eqs. (1)–(3); Section 3.2 Unsupervised losses]
    "We first briefly recap the main ideas of χ [30]… By carefully designed losses that enforce left-right consistency between χM and ¯χM and certain geometrical constraints… To train our model, we use a combination of 3D geometric losses… L=λ1 Ldis +λ2 Lvar +λ3 Lfif +λ4 Limg, where Ldis, Lvar and Lfif are as defined in Eq. (1), Eq. (2) and Eq. (3)"

    The three geometric losses that supply the 3D prior are taken verbatim from the authors’ own prior paper χ. This is a self-citation of the training objectives. It is not load-bearing for the paper’s main claim: the losses are unsupervised consistency/smoothness/balance regularizers (not uniqueness theorems or fitted parameters), the novel hybrid image path and Limg are new, and all reported accuracies are measured on independent annotations rather than being forced by those losses.

full rationale

The derivation chain is not circular. Pix2LR is trained without left-right ground truth: geometric losses (Ldis, Lvar, Lfif) enforce consistency between original and horizontally flipped predictions after surface aggregation, and Limg does the same on masked in-the-wild pixels. These are regularizers, not fits of free parameters to the quantities later reported as predictions. Evaluation uses independently obtained dense labels (from BeCoS vertex GT projected to rendered pixels) and newly hand-labeled sparse keypoint left-right annotations on PF-Pascal/SPair-71k; the accuracy metric’s max-over-swap is a fair orientation-invariant score, not a tautology. The central empirical claim—that a medium-scale human/quadruped 3D prior plus diverse images yields pixel-wise left-right that generalizes to cars, trains, paintings, etc.—is tested against external benchmarks and ablations (Tables 2–5, Figs. 5–6), not forced by construction. The only mild self-citation is reuse of geometric losses and the chirality-subspace idea from the authors’ prior χ work; that is ordinary methodological scaffolding, not a uniqueness theorem or a fitted input renamed as prediction. Score 1 reflects that minor self-citation without elevating it to load-bearing circularity.

Axiom & Free-Parameter Ledger

2 free parameters · 3 axioms · 1 invented entities

The central claim rests on three domain assumptions about foundation-model features and on a small set of loss-weight free parameters; no new physical entities are postulated.

free parameters (2)
  • λ1, λ2, λ3, λ4 (loss weights)
    Relative weights of the four unsupervised losses; chosen by the authors and not derived from first principles.
  • Nr (number of rendered views)
    Number of virtual cameras used for surface aggregation; a design choice that affects feature quality.
axioms (3)
  • domain assumption Vision foundation models (SD, DINOv3) already contain rich, transferable left-right semantic information.
    Stated as the operating hypothesis in Section 2 and used throughout the method; supported by citations but not proved inside the paper.
  • ad hoc to paper A medium-scale collection of only human- and quadruped-animal meshes supplies a sufficient geometric prior for arbitrary object categories.
    Core hybrid-training hypothesis of Section 3.2; empirically tested but not theoretically justified.
  • domain assumption Horizontal flip of an image is equivalent (up to in-plane rotation) to a 3D reflection that swaps left and right.
    Used to construct the flipped feature path; standard projective geometry but crucial for the consistency losses.
invented entities (1)
  • Pix2LR predictor + order-swapped aggregation no independent evidence
    purpose: Enables joint unsupervised training on 3D meshes and wild images by predicting left-right scores before surface aggregation.
    New architectural arrangement relative to χ; no independent physical existence outside the method.

pith-pipeline@v1.1.0-grok45 · 19097 in / 2302 out tokens · 21625 ms · 2026-07-11T10:12:10.723699+00:00 · methodology

0 comments
read the original abstract

While various works address reflective symmetry understanding in 3D data and images, pixel-level semantic left-right prediction of in-the-wild images remains challenging, due to certain difficulties including the lack of 3D information, occlusion, object pose variation, partiality, etc. In this work, we propose an unsupervised learning framework to tackle this challenge. Leveraging recent advances in vertex-wise semantic left-right understanding of 3D data, our unsupervised learning method jointly utilises 3D shape and image datasets to infer pixel-wise semantic left-right predictions in single-view images. In particular, we show that a medium-scale 3D shape dataset comprising mainly of human- and quadruped animal-like shapes, combined with diverse in-the-wild image data, are sufficient to achieve high-quality semantic left-right prediction in images, even for entirely unseen 3D object categories, such as cars or trains. Overall, our approach achieves superior performance in dense pixel-wise semantic left-right predictions on both rendered and in-the-wild image datasets when compared to existing state-of-the-art methods.

Figures

Figures reproduced from arXiv: 2607.05006 by Florian Bernard, Tobias Wei{\ss}berg, Weikang Wang.

Figure 1
Figure 1. Figure 1: We propose the first unsupervised pixel-level semantic left-right prediction framework for in-the-wild images, which is robust across diverse challenging settings. Abstract. While various works address reflective symmetry understand￾ing in 3D data and images, pixel-level semantic left-right prediction of in-the-wild images remains challenging, due to certain difficulties includ￾ing the lack of 3D informati… view at source ↗
Figure 2
Figure 2. Figure 2: Overview of our pixel-wise semantic left-right prediction framework Pix2LR. We use a hybrid training dataset comprising of a medium-scale 3D shape dataset (mainly of human- and animal-like shapes), which we use to render images, together with an in-the-wild image dataset (of a broad class of object categories, e.g., including cars or aeroplanes). Rendered and in-the-wild images are both flipped horizontall… view at source ↗
Figure 3
Figure 3. Figure 3: (a) Given an image I, we use a shared vision foundation model to extract features from both I and its horizontally flipped counterpart I¯. The feature obtained from I¯ is then horizontally flipped back to align with I, yielding the final pair (F, F¯). (b) Solid and dashed lines denote the prediction paths of rendered and in-the-wild images, respectively. The top branch illustrates χ [30] and the bottom bra… view at source ↗
Figure 4
Figure 4. Figure 4: gives examples of these newly introduced keypoint-wise left-right annota￾tions. More details about the annotation are in the supplementary materials [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Qualitative comparisons of dense pixel-level left-right predictions for χBeCoS [30] and OursBeCoS-Both for different object categories sampled randomly from both SPair￾71k [17] and PF-Pascal [7]. Our predictions are significantly more accurate. 4.3 Generalisation Analysis We analyse our trained framework in the following multiple levels to show its strong generalisation ability. Generalisation on 3D inform… view at source ↗
Figure 6
Figure 6. Figure 6: Zero-shot performance of our method on paintings or anime images (top), and on unseen categories neither part of the 3D shape nor the image dataset (bottom). from different categories with large pose and viewpoint variations, as well as occlusion and truncation. Yet, our framework can still performs well on most of these cases, which shows generalisation on images with large geometry variations. Generalisa… view at source ↗
Figure 7
Figure 7. Figure 7: We identify a few challenges for method leading to wrong predictions: (i) multi￾ple symmetries, (ii) out-of-distribution views , (iii) unusual poses, (iv) rare geometries. 5 Discussion & Limitations Overall, our Pix2LR shows superiority compared to other state-of-the-art meth￾ods for semantic left-right understanding of images across a broad range of cat￾egories. Nevertheless, there is room for improvement… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

38 extracted references · 3 linked inside Pith

  1. [1]

    arXiv preprint arXiv:2503.02660 (2025)

    Aguirre, I., Sipiran, I., Montanana, G.: A dataset-free approach for self-supervised learning of 3d reflectional symmetries. arXiv preprint arXiv:2503.02660 (2025)

    Pith/arXiv arXiv 2025

  2. [2]

    In: Proceedings of the IEEE conference on computer vision and pattern recognition

    Bogo, F., Romero, J., Loper, M., Black, M.J.: Faust: Dataset and evaluation for 3d mesh registration. In: Proceedings of the IEEE conference on computer vision and pattern recognition. pp. 3794–3801 (2014)

    2014

  3. [3]

    In: European Conference on Computer Vi- sion

    Cheng, X., Deng, C., Harley, A.W., Zhu, Y., Guibas, L.: Zero-shot image feature consensus with deep functional maps. In: European Conference on Computer Vi- sion. pp. 277–293. Springer (2024)

    2024

  4. [4]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Dutt, N.S., Muralikrishnan, S., Mitra, N.J.: Diffusion 3d features (diff3f): Deco- rating untextured shapes with distilled semantic features. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 4494– 4504 (2024)

    2024

  5. [5]

    In: Computer Graphics Forum

    Ehm, V., Amrani, N.E., Xie, Y., Bastian, L., Gao, M., Wang, W., Sang, L., Cao, D., Weißberg, T., Lähner, Z., et al.: Beyond complete shapes: A benchmark for quantitative evaluation of 3d shape surface matching algorithms. In: Computer Graphics Forum. vol. 44, p. e70186. Wiley Online Library (2025)

    2025

  6. [6]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    El Banani, M., Raj, A., Maninis, K.K., Kar, A., Li, Y., Rubinstein, M., Sun, D., Guibas,L.,Johnson,J.,Jampani,V.:Probingthe3dawarenessofvisualfoundation models. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21795–21806 (2024)

    2024

  7. [7]

    In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2016)

    Ham, B., Cho, M., Schmid, C., Ponce, J.: Proposal flow. In: Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (2016)

    2016

  8. [8]

    In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision

    Hartwig, R., Muhle, D., Marin, R., Cremers, D.: Geco: Geometrically consistent embedding with lightspeed inference. In: Proceedings of the IEEE/CVF Interna- tional Conference on Computer Vision. pp. 9309–9319 (2025)

    2025

  9. [9]

    In: SIGGRAPH Asia 2024 Conference Papers

    Je, J., Liu, J., Yang, G., Deng, B., Cai, S., Wetzstein, G., Litany, O., Guibas, L.: Robust symmetry detection via riemannian langevin dynamics. In: SIGGRAPH Asia 2024 Conference Papers. pp. 1–11 (2024)

    2024

  10. [10]

    In: Computer graphics forum

    Kim, V.G., Lipman, Y., Chen, X., Funkhouser, T.: Möbius transformations for global intrinsic symmetry analysis. In: Computer graphics forum. vol. 29, pp. 1689–

  11. [11]

    Wiley Online Library (2010)

    2010

  12. [12]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A.C., Lo, W.Y., et al.: Segment anything. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 4015–4026 (2023)

    2023

  13. [13]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Li, R.W., Zhang, L.X., Li, C., Lai, Y.K., Gao, L.: E3sym: Leveraging e (3) invari- ance for unsupervised 3d planar reflective symmetry detection. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 14543–14553 (2023)

    2023

  14. [14]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Li, X., Huang, Z., Thai, A., Rehg, J.M.: Symmetry strikes back: From single-image symmetry detection to 3d generation. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 743–752 (2025) 16 Weikang Wang et al

    2025

  15. [15]

    In: Computer Graphics Forum

    Liu, T., Kim, V.G., Funkhouser, T.: Finding surface correspondences using sym- metry axis curves. In: Computer Graphics Forum. vol. 31, pp. 1607–1616. Wiley Online Library (2012)

    2012

  16. [16]

    Foundations and Trends®in Computer Graphics and Vision5(1–2), 1–195 (2010)

    Liu, Y., Hel-Or, H., Kaplan, C.S., Van Gool, L., et al.: Computational symmetry in computer vision and computer graphics. Foundations and Trends®in Computer Graphics and Vision5(1–2), 1–195 (2010)

    2010

  17. [17]

    Advances in Neural Information Processing Systems38, 15841–15868 (2026)

    Mariotti, O., Du, Z., Bhalgat, Y., Mac Aodha, O., Bilen, H.: Jamais vu: Exposing the generalization gap in supervised semantic correspondence. Advances in Neural Information Processing Systems38, 15841–15868 (2026)

    2026

  18. [18]

    arXiv prepreint arXiv:1908.10543 (2019)

    Min, J., Lee, J., Ponce, J., Cho, M.: Spair-71k: A large-scale benchmark for se- mantic correspondence. arXiv prepreint arXiv:1908.10543 (2019)

    Pith/arXiv arXiv 1908

  19. [19]

    ACM Transactions on Graphics (ToG)25(3), 560–568 (2006)

    Mitra, N.J., Guibas, L.J., Pauly, M.: Partial and approximate symmetry detection for 3d geometry. ACM Transactions on Graphics (ToG)25(3), 560–568 (2006)

    2006

  20. [20]

    In: Computer graphics forum

    Mitra, N.J., Pauly, M., Wand, M., Ceylan, D.: Symmetry in 3d geometry: Ex- traction and applications. In: Computer graphics forum. vol. 32, pp. 1–23. Wiley Online Library (2013)

    2013

  21. [21]

    In: Pro- ceedings of the European Conference on Computer Vision (ECCV)

    Nagar, R., Raman, S.: Fast and accurate intrinsic symmetry detection. In: Pro- ceedings of the European Conference on Computer Vision (ECCV). pp. 417–434 (2018)

    2018

  22. [22]

    Transactions on Machine Learning Research Journal (2024)

    Oquab, M., Darcet, T., Moutakanni, T., Vo, H., Szafraniec, M., Khalidov, V., Fernandez, P., Haziza, D., Massa, F., El-Nouby, A., et al.: Dinov2: Learning robust visual features without supervision. Transactions on Machine Learning Research Journal (2024)

    2024

  23. [23]

    In: Computer graphics forum

    Ovsjanikov, M., Sun, J., Guibas, L.: Global intrinsic symmetries of shapes. In: Computer graphics forum. vol. 27, pp. 1341–1348. Wiley Online Library (2008)

    2008

  24. [24]

    IEEE Transactions on Visualization and Computer Graphics29(9), 3799–3808 (2022)

    Qiao, Y.L., Gao, L., Liu, S.Z., Liu, L., Lai, Y.K., Chen, X.: Learning-based in- trinsic reflectional symmetry detection. IEEE Transactions on Visualization and Computer Graphics29(9), 3799–3808 (2022)

    2022

  25. [25]

    In: International conference on machine learning

    Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J., et al.: Learning transferable visual models from natural language supervision. In: International conference on machine learning. pp. 8748–8763. PmLR (2021)

    2021

  26. [26]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Rombach, R., Blattmann, A., Lorenz, D., Esser, P., Ommer, B.: High-resolution image synthesis with latent diffusion models. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 10684–10695 (2022)

    2022

  27. [27]

    In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition

    Seo, A., Kim, B., Kwak, S., Cho, M.: Reflection and rotation symmetry detection via equivariant learning. In: Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. pp. 9539–9548 (2022)

    2022

  28. [28]

    In: Proceedings of the IEEE/CVF international conference on computer vision

    Seo, A., Shim, W., Cho, M.: Learning to discover reflection symmetry via polar matching convolution. In: Proceedings of the IEEE/CVF international conference on computer vision. pp. 1285–1294 (2021)

    2021

  29. [29]

    arXiv preprint arXiv:2508.10104 (2025)

    Siméoni, O., Vo, H.V., Seitzer, M., Baldassarre, F., Oquab, M., Jose, C., Khali- dov, V., Szafraniec, M., Yi, S., Ramamonjisoa, M., et al.: Dinov3. arXiv preprint arXiv:2508.10104 (2025)

    Pith/arXiv arXiv 2025

  30. [30]

    In: Proceedings of the Computer Vision and Pattern Recognition Conference

    Wandel, K., Wang, H.: Semalign3d: Semantic correspondence between rgb-images through aligning 3d object-class representations. In: Proceedings of the Computer Vision and Pattern Recognition Conference. pp. 1138–1147 (2025)

    2025

  31. [31]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Wang, W., Weißberg, T., El Amrani, N., Bernard, F.: kh: Symmetry understand- ing of 3d shapes via chirality disentanglement. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 28292–28302 (2025) Pix2LR 17

    2025

  32. [32]

    IEEE Transactions on Image Processing (2024)

    Wang, Y., Luo, C.: Key-axis-based localization of symmetry axes in 3d objects utilizing geometry and texture. IEEE Transactions on Image Processing (2024)

    2024

  33. [33]

    In: Thirteenth In- ternational Conference on 3D Vision (2026)

    Weißberg, T., Wang, W., Roetzer, P., El Amrani, N., Bernard, F.: Symmetry in- formative and agnostic feature disentanglement for 3d shapes. In: Thirteenth In- ternational Conference on 3D Vision (2026)

    2026

  34. [34]

    In:Proceedings ofthe IEEE/CVF International Conferenceon Computer Vision

    Yang, T., Rahman, M.A., Yeh, R.A.: Clipsym: Delving into symmetry detection with clip. In:Proceedings ofthe IEEE/CVF International Conferenceon Computer Vision. pp. 21003–21013 (2025)

    2025

  35. [35]

    In: Proceedings of the IEEE/CVF International Conference on Computer Vision

    Yu, W., Seo, A., Cho, M.: Axis-level symmetry detection with group-equivariant representation. In: Proceedings of the IEEE/CVF International Conference on Computer Vision. pp. 24791–24800 (2025)

    2025

  36. [36]

    In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition

    Zhang, J., Herrmann, C., Hur, J., Chen, E., Jampani, V., Sun, D., Yang, M.H.: Telling left from right: Identifying geometry-aware semantic correspondence. In: ProceedingsoftheIEEE/CVFConferenceonComputerVisionandPatternRecog- nition. pp. 3076–3085 (2024)

    2024

  37. [37]

    In: Advances in Neural Information Processing Systems

    Zhang, J., Herrmann, C., Hur, J., Polania Cabrera, L., Jampani, V., Sun, D., Yang, M.H.: A tale of two features: Stable diffusion complements dino for zero-shot semantic correspondence. In: Advances in Neural Information Processing Systems. vol. 36, pp. 45533–45547 (2023)

    2023

  38. [38]

    In: International Conference on Learning Representations

    Zhu, J., Ju, Y., Zhang, J., Wang, M., Yuan, Z., Hu, K., Xu, H.: Densematcher: Learning 3d semantic correspondence for category-level manipulation from a single demo. In: International Conference on Learning Representations. vol. 2025, pp. 49538–49559 (2025)

    2025