Paper deep dive
KASALv2: Fully Automatic 3D Rotational Symmetry Classification and Axis Localization
Mengxin Zhang, Yulin Wang, Chen Luo, Yongzhe Li, Yijun Zhou
Intelligence
Status: succeeded | Model: Gemma-4-26B-A4B | Prompt: intel-v1 | Confidence: 92%
Last extracted: 10/11/2026, 2:40:42 AM
Summary
The paper introduces KASALv2, a fully automatic, reference-free framework for 3D rotational symmetry classification and axis localization. It utilizes a Geometric Degeneration Hierarchy (GDH) to localize dominant high-order axes, infer rotational orders via self-consistency analysis, and reconstruct full symmetry structures across all eight canonical types. A texture-aware extension handles appearance-induced symmetry reductions. Experiments show 94.75% accuracy on the GSO dataset and improved 6D pose estimation accuracy when used as priors for FoundationPose.
Entities (12)
Relation Signals (8)
KASALv2 → evaluatedon → GSO
confidence 95% · achieving 94.75% accuracy on 438 symmetric objects in GSO.
KASALv2 → evaluatedon → BOP
confidence 95% · Experiments on... BOP... validate generalization and scalability
KASALv2 → uses → Geometric Degeneration Hierarchy
confidence 95% · Central to the method is the Geometric Degeneration Hierarchy (GDH)...
KASALv2 → improves → 6D Pose Estimation
confidence 90% · Training FoundationPose with these priors improves accuracy by up to 0.9%... showing that automatically estimated rotational priors improve downstream 6D pose estimation.
Geometric Degeneration Hierarchy → includes → A4
confidence 90% · Platonic family... tetrahedral (A4)
Geometric Degeneration Hierarchy → includes → C_inf
confidence 90% · continuous types C∞ and D∞... dihedral–cyclic family... C∞⇒Cn
Geometric Degeneration Hierarchy → includes → A5
confidence 90% · Platonic family... icosahedral (A5)
Geometric Degeneration Hierarchy → includes → S4
confidence 90% · Platonic family... octahedral (S4)
Cypher Suggestions (0)
No Cypher suggestions yet.
Abstract
Abstract:Rotational symmetry is an important prior in 6D pose estimation, improving pose accuracy and supporting symmetry-aware evaluation. However, current symmetry annotations for 3D objects remain largely manual or semi-automatic, often requiring predefined types or orders, which limits scalability. This work introduces a fully automatic, reference-free framework for symmetry-type classification, rotational-order identification, and full-axis localization across all eight canonical 3D rotational symmetry types. The method localizes a dominant high-order axis, infers its rotational order through self-consistency analysis, and reconstructs the complete symmetry structure under a hierarchy-guided formulation. A texture-aware extension further models appearance-induced reductions in rotational order while preserving axis orientations. Experiments on idealized and real-world datasets demonstrate strong accuracy and generalization, achieving 94.75% accuracy on 438 symmetric objects in GSO. Training FoundationPose with these priors improves accuracy by up to 0.9% across five BOP datasets, showing that automatically estimated rotational priors improve downstream 6D pose estimation. Code is available at this https URL.
Tags
Links
- Source: https://arxiv.org/abs/2610.09534v1
- Canonical: https://arxiv.org/abs/2610.09534v1
Trouble viewing inline? Open PDF directly →
Full Text
47,954 characters extracted from source content.
Expand or collapse full text
KASALv2: Fully Automatic 3D Rotational Symmetry Classification and Axis Localization Mengxin Zhang †thanks: Co-first authors (equal contribution). Yulin Wang11footnotemark: 1 Chen Luo †thanks: Co-corresponding authors. Yongzhe Li Yijun Zhou22footnotemark: 2 Affiliation: School of Mechanical Engineering, Southeast University Affiliation: Nanjing, China Affiliation: mx.zhang, yulinwang, chenluo, yongzheli, zhouyj@seu.edu.cn Abstract Rotational symmetry is an important prior in 6D pose estimation, improving pose accuracy and supporting symmetry-aware evaluation. However, current symmetry annotations for 3D objects remain largely manual or semi-automatic, often requiring predefined types or orders, which limits scalability. This work introduces a fully automatic, reference-free framework for symmetry-type classification, rotational-order identification, and full-axis localization across all eight canonical 3D rotational symmetry types. The method localizes a dominant high-order axis, infers its rotational order through self-consistency analysis, and reconstructs the complete symmetry structure under a hierarchy-guided formulation. A texture-aware extension further models appearance-induced reductions in rotational order while preserving axis orientations. Experiments on idealized and real-world datasets demonstrate strong accuracy and generalization, achieving 94.75% accuracy on 438 symmetric objects in GSO. Training FoundationPose with these priors improves accuracy by up to 0.9% across five BOP datasets, showing that automatically estimated rotational priors improve downstream 6D pose estimation. Code is available at https://github.com/WangYuLin-SEU/KASAL. 1 Introduction Rotational symmetry is a key geometric prior in 6D pose estimation[1, 2, 3, 4, 5, 6, 7]. It resolves pose ambiguity, stabilizes optimization, and supports symmetry-aware metrics such as ADD-S[8], MSSD, and MSPD[9]. Recent systems showing HccePose(BF)[10] and ZebraPose[11] use symmetry priors to improve accuracy, indicating that reliable symmetry information benefits high-precision estimation. Meanwhile, the field is moving toward unseen-object generalization and large-scale synthetic training. These trends rely on datasets with tens of thousands of 3D models, where per-object rotational symmetry is needed but manual annotation is infeasible. Such scalable symmetry annotation also benefits mesh generation and large-scale 3D asset processing. Figure 1: Comparative landscape of rotational symmetry analysis methods categorized by geometric reference dependency and task capability. The proposed method supports all eight canonical types without geometric references and handles texture-induced symmetries. In addition, rotational symmetry is structurally diverse, spanning eight canonical forms[12, 13, 14, 15, 16] with distinct axis multiplicities, rotational orders, and geometric configurations. This diversity produces many subtypes and requires a unified framework capable of joint reasoning over type and axis structure. However, as shown in Fig. 1, existing methods remain limited in automation and coverage: many address only subsets of symmetry types,[17, 18, 19, 20, 21, 22, 23] others depend on geometric references,[24, 25] and recent all-type approaches[26] still require manual specification of rotational orders or categories. For large-scale unseen-object pose estimation, an automatic system capable of full symmetry classification and axis localization is needed. To address this requirement, we extend the key-axis-based localization method of Wang et al. [26], which relies on manually specified symmetry types and orders, and present KASALv2, a unified framework for complete rotational-symmetry inference across all eight canonical symmetry types without any prior input or predefined geometric references. KASALv2 transforms the task from type-conditioned axis localization into fully automatic analysis that jointly determines the symmetry type, rotational order, and the full set of canonical axes. Beyond removing manual input, it replaces the exhaustive directional enumeration of [26] with the Geometric Degeneration Hierarchy (GDH), a computable reasoning mechanism that encodes type-specific geometric relations and guides the symmetry analysis. Guided by GDH, the system localizes a dominant high-order axis, estimates its rotational order through periodic self-consistency, infers the corresponding symmetry family, and reconstructs all canonical axes in a unified manner, with a texture-aware extension that accounts for appearance-induced reductions in rotational order. Extensive experiments on DSRSTO[26], BOP[27], and GSO[28] validate generalization and scalability, achieving 94.75% accuracy on real-world symmetric objects and demonstrating robust performance on the DSRSTO texture subset under appearance-induced degradation. When used as symmetry priors for FoundationPose,[29] the method improves BOP accuracy by up to 0.9%, confirming the value of precise symmetry estimation for downstream pose prediction. Figure 2: The pipeline integrates a Geometric Degeneration Hierarchy (GDH) and an operational framework. The GDH establishes the hierarchical taxonomy of 3D rotational symmetry, while the operational framework implements it. Given a 3D object of unknown symmetry, the system first localizes a dominant high-order axis and infers its rotational order through self-consistency optimization. Based on this order, candidate symmetry types are derived and verified by checking whether a secondary axis satisfies the corresponding geometric constraints. This hierarchical verification yields type classification and reconstruction of all canonical axes under the identified geometric relations. Finally, the texture-symmetry module refines the classification by modeling appearance-induced degradation, covering reduced, invariant, and discretized texture cases. Our contributions are summarized as follows: • A unified and fully automatic framework for 3D rotational symmetry analysis that performs symmetry-type classification, rotational-order identification, and full-axis localization across all eight canonical types in a reference-free pipeline. • A geometry-driven formulation, the Geometric Degeneration Hierarchy (GDH), introduced as a computable geometric hierarchy that organizes rotational symmetries via structured axis reduction and provides the algorithmic pathway for mapping an unknown object to its canonical symmetry family. • An operational instantiation of GDH enabling consistent type determination, stable high-order axis localization, and a texture-aware extension that models appearance-induced reductions in rotational order while preserving axis orientations. • A scalable and annotation-free source of rotational symmetry priors for unseen-object pose estimation. Training FoundationPose with these priors on a one-million-image corpus yields up to a 0.9% gain on BOP, highlighting the value of symmetry-aware augmentation for foundation-level models. 2 Related Work Rotational symmetry analysis in 3D vision comprises two fundamental tasks: symmetry-type identification[24, 25, 17, 18, 23], which infers the underlying rotational structure from data, and symmetry-axis localization[19, 20, 21, 22, 26], which assumes the type is known and estimates its corresponding axes. Existing approaches addressing one or both tasks vary mainly along two design dimensions: (i) reliance on geometric priors such as centroids, keypoints, or object frames, and (i) coverage of the eight canonical rotational symmetry types. These dimensions strongly influence robustness and generality. 2.1 Symmetry Type Identification Symmetry type detection identifies whether an object exhibits rotational symmetry and, if so, infers its specific type and order. Existing approaches fall into two categories: reference-based and reference-free. Reference-based methods [24, 25] rely on geometric anchors such as centroids, midpoints, or principal axes. They perform well under controlled conditions but are highly sensitive to surface noise, occlusion, and misalignment, which limits their applicability in real-world 6D pose estimation. Reference-free methods [17, 18, 23] infer symmetry directly from surface geometry through optimization, spectral analysis, or voting. They achieve greater robustness by avoiding predefined anchors but still lack generality across all canonical rotational types. For example, Hruda et al.[18] proposed a reflection-based framework that fails on continuously symmetric objects such as cylinders or cones, while Rajkumar et al.[17] presented a hybrid approach limited to continuous and second-order discrete conical symmetries. 2.2 Symmetry Axis Localization Most existing methods address only limited cases[19, 20, 21, 22, 30] and are typically designed for specific scenarios. For example, Sipiran[20, 21] proposed approaches for cultural heritage restoration that focuses on continuous types. These methods, however, cannot accommodate texture-induced symmetries. Although Wang et al.[26] introduced a method capable of handling all symmetry types, including texture-related cases, it still requires manual specification of symmetry type and order. This dependency hinders full automation and leads to annotation ambiguity, particularly for objects with high-order or composite symmetries that even experts may find difficult to classify accurately. 2.3 Rotational Symmetry in Pose Estimation In 6D pose estimation, existing methods for handling rotationally symmetric objects fall into two categories. The first does not use explicit symmetry priors[3, 4, 31, 32] but incorporates architectural or learning designs that implicitly capture latent symmetry cues. Examples include EPOS[32] and GDR-Net[31]. The second[1, 7, 2, 33, 34, 11, 10] explicitly leverages known symmetry priors to resolve pose ambiguities and improve accuracy. Methods such as HccePose (BF)[10] and ZebraPose[11] use predefined symmetry matrices to transform ground-truth rotations, selecting the variant closest to the identity as the training target. This strategy reduces symmetry-induced ambiguity and leads to more stable training and higher accuracy than prior-free methods. Symmetry priors also underpin evaluation protocols. Metrics such as ADD-S[8], MSSD[9], and MSPD[9] define pose equivalence based on symmetry-aware criteria; without them, physically identical yet numerically distinct poses would be misjudged. Hence, symmetry priors are essential not only for accurate estimation but also for consistent evaluation. Despite their importance, existing symmetry annotations remain largely manual or require predefined types or orders, limiting scalability. These limitations motivate the development of a fully automatic framework for robust object symmetry analysis without human intervention. To this end, KASALv2 is introduced, enabling automatic detection of symmetry types and precise axis localization across all eight canonical rotational types, thereby providing reliable symmetry priors for downstream pose estimation. 3 Methodology Our framework performs reference-free rotational symmetry analysis for 3D objects. Central to the method is the Geometric Degeneration Hierarchy (GDH), a computable structure encoding all canonical rotational symmetry families and the degeneration conditions between them. Guided by GDH, the method localizes the dominant high-order axis, estimates its order, determines the symmetry type, and reconstructs the full canonical axis configuration. A texture-aware extension further models appearance-induced order reductions while preserving geometric axis orientations. The overall workflow is shown in Fig. 2. This section follows a theory-to-operation progression. Sec. 3.1 introduces GDH, which defines the structural relations and degeneration pathways among all rotational symmetry families. Sec. 3.2 and Sec. 3.3 then apply these relations to localize the dominant axis, infer its order, classify the symmetry type, and recover all canonical axes. Sec. 3.4 extends the framework to appearance-aware settings using texture-induced order reductions. 3.1 Geometric Degeneration Hierarchy (GDH) for 3D Rotational Symmetry Rotational symmetry in three-dimensional space corresponds to the orientation-preserving subgroups of SO(3)SO(3), including the continuous types C∞C_∞ and D∞D_∞, the finite cyclic and dihedral groups CnC_n and DnD_n, and the rotational groups of the Platonic solids (A4,S4,A5)(A_4,S_4,A_5) [35, 36, 12, 13, 14, 15, 16]. These families cover all canonical rotational symmetries considered in this work, with classical subgroup embeddings such as A4⊂S4A_4⊂ S_4 and A4⊂A5A_4⊂ A_5 [37, 38]. Building on this foundation, we introduce the Geometric Degeneration Hierarchy (GDH), a geometry-driven and algebraically complete hierarchy that reorganizes these SO(3)SO(3) subgroups into an operational form. Rather than a purely conceptual taxonomy, GDH encodes how axis multiplicity, rotational order, and characteristic inter-axis relations evolve under geometric degenerations, providing computable constraints that restrict feasible symmetry families and support type-specific axis inference in our fully automatic framework. At the top of this hierarchy lies the maximally symmetric sphere, where every direction is a valid rotational axis. From this state, GDH considers two geometric degeneration lineages: the dihedral–cyclic family, characterized by reductions from C∞C_∞ and D∞D_∞ to their finite counterparts, and the Platonic family, representing discrete reductions of multi-axis structures such as A5A_5, S4S_4, and A4A_4. The dihedral–cyclic lineage begins with the cylindrical form D∞D_∞, characterized by continuous rotation about a single principal axis. From this state, symmetry may evolve along two geometric reduction paths. The first is the discretization of continuous rotation, D∞⇒Dn,C∞⇒Cn,n∈ℕ,n≥2,D_∞ D_n, C_∞ C_n, n ,\;n≥ 2, (1) producing finite n-fold rotational structures with or without orthogonal two-fold axes. This discretization relation allows the dominant axis order to restrict the feasible discrete families during type inference. The second path corresponds to the loss of orthogonal axes, D∞⇒C∞,Dn⇒Cn,D_∞ C_∞, D_n C_n, (2) yielding single-axis rotational configurations. In parallel, the Platonic lineage models the geometric reduction of multiple high-order axes present in fully symmetric structures. This leads to the icosahedral (A5)(A_5), octahedral (S4)(S_4), and tetrahedral (A4)(A_4) symmetries, which exhibit progressively fewer and lower-order rotational axes: A5⇒A4,S4⇒A4.A_5 A_4, S_4 A_4. (3) Each downward transition reflects a geometric simplification in both the number and orientation of rotational axes. Together, the dihedral–cyclic and Platonic lineages form the Geometric Degeneration Hierarchy (GDH), a geometry-based organizational framework encompassing all eight canonical types of 3D rotational symmetry. Within this hierarchy, the localized dominant high-order axis provides the primary constraint that narrows the feasible subtype of symmetry types, while its estimated rotational order further restricts the set of candidates. The identification of a secondary axis that satisfies the corresponding geometric relations is then used to verify the specific type. Once the type is determined, its intrinsic axis relations uniquely specify the remaining canonical axes, enabling complete reconstruction of the object’s rotational structure. Through these components, GDH supplies a structured set of geometric constraints that make the subsequent classification and axis reconstruction steps operationally tractable within our fully automatic framework. 3.2 High-order Axis Localization and Order Identification 3.2.1 Dominant Axis Localization Guided by the GDH framework introduced in Sec. 3.1, this stage localizes the dominant high-order rotational axis, which anchors subsequent order identification and type classification. Unlike reference-based methods that depend on pre-aligned templates or centroids, our approach operates entirely in a reference-free manner by optimizing a global alignment objective directly on the unit sphere. Alignment Principle. To locate the true rotational axis without prior knowledge of symmetry order, candidate directions are evaluated using a mixed-order alignment loss that aggregates alignment errors across multiple potential orders: Lmix(a)=∑n∈NLn(a)L_mix (a )= Σ_n∈ NL_n (a )~ (4) Ln(a)=∑k=1n−1Chamfer(P,R(a,θk)P)L_n (a )= k=1 n-1 Σ\,Chamfer (P,R (a,θ_k )P ) (5) where θk=2πk/nθ_k=2π k/n and R(a,θk)R (a,θ_k ) denotes rotation of the point cloud P around axis a by θkθ_k and Chamfer(·,·)( , ) denotes the Chamfer distance[39]. Figure 3: Aggregated alignment-loss distribution on the unit sphere. Low-loss regions (orange) cluster around the true high-order axis, illustrating the self-consistent nature of high-order symmetries. Because high-order axes inherently satisfy the periodicity of lower-order rotations, their aggregated loss tends to be smaller, forming natural descent basins in the loss landscape (Fig. 3). As a result, optimization is naturally biased toward these dominant high-order directions during search. This self-guiding property enables high-order symmetries to emerge automatically, thereby supporting fully reference-free localization of the dominant axis. Two-Stage Sampling and Optimization. To balance global exploration and computational efficiency, axis localization is carried out in two stages: global coarse sampling followed by local fine refinement, with a final gradient-based optimization for convergence stability. In the first stage, global coarse sampling uniformly explores the unit sphere to identify regions of low mixed-order alignment loss, yielding the top-kcandk_cand candidate directions. The second stage performs local fine sampling around each candidate by resampling a spherical cap of angular radius r and re-evaluating top-kinitk_init directions under the same loss. This hierarchical search strategy enables efficient localization of promising minima while maintaining broad coverage of the loss landscape. The refined candidates are then optimized using the Adam algorithm to further minimize the aggregated loss and stabilize the alignment result. Through this progressive sampling-to-optimization pipeline, the framework robustly localizes the dominant high-order rotational axis, which subsequently serves as the geometric anchor for order estimation. 3.2.2 Rotational Order Identification After the dominant axis a*a^* is localized, rotational order NestN_est is inferred by analyzing the object’s periodic self-alignment under densely sampled rotations. The object is rotated around a*a^* through a dense set of uniformly spaced angles ϕ∈[0,2π)φ∈ [0,2π ) (typically 360 samples per full revolution), and the Chamfer distance between the original and rotated shapes is computed at each step to form a rotational self-similarity signal. The dominant frequency of this signal reveals the object’s fundamental rotational periodicity. A coarse estimate NfftN_fft is obtained from the dominant frequency component and further refined through direct alignment evaluation over candidate orders Nfft,Nfft±1 \N_fft,~N_fft± 1 \. The order that minimizes the reconstruction error is selected as the final NestN_est. Ambiguous or degenerate cases are resolved through hierarchical verification: when Nest≥3N_est≥ 3, the object is considered discretely rotationally symmetric; when Nest=2N_est=2, a 180∘180 rotation test distinguishes genuine two-fold symmetry from reflection-like structures; and when Nest≤1N_est≤ 1, a small-angle (45∘45 ) rotation test differentiates continuous from asymmetric shapes. This procedure provides a robust, reference-free estimation of both the dominant axis and its rotational order, forming a reliable basis for the hierarchical symmetry classification and full-axis reconstruction introduced in Sec. 3.3. 3.3 Hierarchical Type Classification and Full-Axis Reconstruction Building upon the Geometric Degeneration Hierarchy (GDH) introduced in Section 3.1, each symmetry type is characterized by the number and relative orientation of its rotational axes. Once the dominant high-order axis a∗a^* is localized and its order NestN_est is inferred, the GDH restricts the hypothesis space to the subset of types consistent with NestN_est. The remaining ambiguity concerns the orientation of a secondary high-order axis relative to a∗a^*. Under GDH, each candidate symmetry type prescribes a characteristic inter-axis inclination αF _F between a∗a^* and a secondary axis b∗b^*, which may share the same or a lower rotational order. Because these inclinations are invariant to pose and scale, verifying the existence of a secondary axis at the prescribed αF _F serves to identify the correct symmetry type and resolve the remaining ambiguity. 3.3.1 Secondary-Axis Identification and Hierarchical Type Classification The secondary axis b∗b^* is searched along a circular locus on the unit sphere, defined by a fixed inclination αFα_F from the dominant axis a∗a^*, as prescribed by the corresponding GDH type. The optimal direction is obtained by minimizing the alignment loss b*=argminb∈S( α F)LChamfer(b,N)b^*= \,b∈ S ( \!\!α\!\! _F ) \,L_Chamfer (b;N ) (6) where N denotes the hypothesized order of b (equal to NestN_est for Platonic families or a lower order for prismatic and pyramidal ones). Candidate directions with the lowest losses are preserved as top-k seeds. Each seed undergoes a one-degree-of-freedom gradient refinement along the ring parameter axis, inherently enforcing the inclination constraint. The refinement, performed using the Adam optimizer, minimizes LChamfer(b,N)L_Chamfer (b;N ). Once b*b^* is obtained, the symmetry type is determined from the order pair (Na*,Nb*) (N_a^*,N_b^* ) and the theoretical inclination αFα_F defined in the GDH. Continuous symmetries (C∞,D∞) (C_∞,D_∞ ) are recognized when no discrete periodicity is found along a*a^*. The presence of additional perpendicular two-fold axes distinguishes cylindrical (D∞D_∞) from circular (C∞C_∞) symmetry, while complete isotropy corresponds to the spherical case. Discrete symmetries (CnC_n, DnD_n) are identified when Nest≥2N_est≥ 2. Orthogonal two-fold axes indicate dihedral (DnD_n) structures, whereas their absence yields cyclic (CnC_n) symmetry. Platonic types (A4A_4, S4S_4, A5A_5) are verified by comparing the measured inter-axis inclination with the theoretical constants 70.53∘70.53 (tetrahedral, A4A_4), 90∘90 (octahedral/cubic, S4S_4), and 63.43∘63.43 (icosahedral, A5A_5). This geometric classification maps the detected axis configuration to its canonical rotational type, completing symmetry-type determination before the full-axis reconstruction in Sec. 3.3.2. 3.3.2 Canonical Axis Reconstruction Once the symmetry type is determined, the GDH provides its canonical axis configuration. With the dominant axis a∗a^* and secondary axis b∗b^* already localized, the remaining axes are recovered by aligning the canonical arrangement associated with the identified type. Each type includes a predefined axis template specifying the relative orientations and multiplicities of its rotational axes. A rigid transformation Q is estimated to align the template pair (a^,b^)( a, b) with the detected pair (a∗,b∗)(a^*,b^*), yielding the full set of axes as uifinal=Qui^.u_i^final=Q\, u_i. (7) This alignment transfers type-specific geometric relations to the object, ensuring consistency with its rotational structure. The resulting reconstruction completes full-axis localization in a compact, reference-free form suitable for symmetry reasoning and pose evaluation. 3.4 Texture-Symmetry Extension for Appearance-Aware Analysis Real-world objects rarely exhibit perfect geometric regularity. Surface appearance, including texture and color patterns, can break rotational invariance in the image domain. To model this effect, texture is treated as a secondary order reduction applied on top of geometric symmetry. Following the GDH, this reduction alters only the rotational order of each axis while preserving its orientation. The effective orders must lie in the divisor set of the geometric order: ntex∈d∣d divides ngeo∪1.n_tex∈\\,d d divides n_geo\,\∪\1\. (8) Thus, a geometrically six-fold object may appear three- or two-fold under periodic textures, and continuous symmetries such as cylinders can manifest as discrete DnD_n patterns when repeated surface structure is present. Given the recovered geometric type and axis configuration, texture refinement is computed analytically by evaluating appearance consistency around each axis, yielding a unified representation that incorporates both geometry- and texture-induced symmetry. 4 Experiments This section presents the experimental setup, ablation studies, and cross-dataset evaluations. We first outline the implementation and evaluation settings, then conduct ablations on the DSRSTO [26] dataset. To assess generalization, we further evaluate on GSO[28] and BOP[27], which include diverse real-world objects with varying symmetry characteristics. 4.1 Experimental Setup 4.1.1 Implementation Details All experiments are implemented in Python/PyTorch and executed on an Intel i5-13400F / RTX 3050 GPU and 32 GB RAM. In the high-order axis detection stage, 128 coarse directions are uniformly sampled on the upper hemisphere. The top six are refined within 10∘10 spherical caps using eight local samples, and the best three undergo Adam optimization (lr=0.04, 5 steps). The mixed-order alignment loss uses the candidate order set N=3,4,5,6N= \3,4,5,6 \. For secondary-axis localization, Adam is applied with lr=0.05 for 5 steps. The number of top candidates equals twice the estimated rotational order, and each region is refined by eight local samples. 4.1.2 Adaptive Sampling Mechanism High-Order Axis Detection. A single spherical-cap radius r governs all sampling budgets within the valid range r∈[rmin,rmax]r∈ [r_min,r_max ], where coverage remains geometrically meaningful. The theoretical lower bound of the coarse-sampling count is derived from the ratio between the sphere surface and the cap area, Nmin=21−cosrN_min= 21- r (9) representing the minimal number of caps required for full coverage. Sampling is limited to the upper hemisphere to avoid antipodal redundancy, and NminN_min is rounded to the nearest multiple of eight for computational convenience. The radius r jointly determines the coarse-sampling count Nmin(r)N_min (r ), candidate directions kcand(r)k_cand (r ), local fine samples nlocal(r)≥4n_local (r )≥ 4, and Adam initializations kinit(r)k_init (r ). Smaller r increases kcand(r)k_cand (r ) to capture narrow high-order axes, while larger r raises kinit(r)k_init (r ) for broader exploration. Both are capped at six, consistent with the fact that in 3D finite rotational symmetries the number of maximal-order (≥3)(≥ 3) axes never exceeds six (achieved only by the icosahedral case). This radius-driven formulation balances sampling density and computational efficiency, ensuring stable localization across geometries. Secondary-Axis Localization. Sampling density scales linearly with the target rotational order, allocating proportionally more candidates and local samples for higher orders to maintain consistent angular resolution. 4.1.3 Metrics and Datasets To quantitatively evaluate the proposed framework, both classification and localization metrics are employed. Classification Metrics. To assess symmetry-type and order prediction, we report three accuracies: acct acc_t =NcorrecttypeNall, = N_correct\ typeN_all, (10a) acco acc_o =Ncorrectorder|correcttypeNcorrecttype, = N_correct\ order\ |\ correct\ typeN_correct\ type, (10b) accT acc_T =Ncorrecttype&orderNall. = N_correct\ type\ \&\ orderN_all. (10c) These metrics respectively reflect type accuracy, order-consistent accuracy, and overall correctness. Localization Metric. Axis localization is evaluated using the Average Distance of Model Points (ADI), which measures the discrepancy between the original model and its reconstructed symmetric instances. Given the predicted symmetry orders, directions, and centers, the rotational symmetry is represented by a discrete rotation set Tp=TiT_p=T_i. The localization error is eADI=∑Ti∈TpADI(Ti,I)|Tp|,e_ADI= _T_i∈ T_pADI(T_i,I)|T_p|, (11) where ADI(Ti,I)ADI(T_i,I) denotes the mean pointwise distance between the model and its rotated counterpart. Misclassified objects yield incorrect canonical axis groups and thus invalid rotation sets, making eADI/de_ADI/d meaningless. Accordingly, eADI/de_ADI/d is reported only for samples correctly identified in both type and order. Datasets. The DSRSTO[26] dataset includes 38 CAD objects across seven categories, covering all canonical 3D rotational symmetry types with 27 geometrically symmetric and 11 texture-symmetric models. As a manually designed benchmark, it serves to evaluate symmetry classification and axis localization. To assess performance in real scenarios, tests are conducted on common 6D pose benchmarks such as ITODD[40], T-LESS[41], LM-O[42], HB[43], YCB-V[8], and IC-BIN[44], which contain real-world objects but limited symmetry diversity. The GSO[28] dataset extends the evaluation to 944 scanned models, among which 438 exhibit rotational symmetry, providing a large-scale testbed for generalization across geometries. 4.2 Ablation Studies Ablation experiments on the DSRSTO dataset were conducted to analyze the influence of core design choices on classification accuracy, axis-localization precision, and computational efficiency. The evaluation considers overall accuracy accTacc_T, normalized geometric alignment error eADI/de_ADI/d (where d is the object diameter), and average runtime per object. Two factors were examined: the adaptive radius-driven sampling configuration and the composition of the candidate order set. Figure 4: Effect of the spherical-cap radius r on sampling performance. Dashed curves indicate fitted trends for total accuracy (blue) and normalized alignment error eADI/de_ADI/d (orange). The optimal configuration appears near r=10∘r=10 . Table 1: Ablation experiments on candidate order sets N. Order Set N accTacc_T (%) eADI/de_ADI/d Time (s) 3,4\3,4\ 70.37 0.00419 1.02 3,4,5\3,4,5\ 77.78 0.00423 1.20 3,5,6\3,5,6\ 77.78 0.00411 1.69 3,4,5,6\3,4,5,6\ 81.48 0.00418 1.46 3,4,5,6,7\3,4,5,6,7\ 81.48 0.00419 1.77 3,4,5,6,7,9\3,4,5,6,7,9\ 81.48 0.00429 2.40 2,3,4,5,6\2,3,4,5,6\ 77.78 0.00659 2.28 4.2.1 Sampling Configuration The effect of the spherical-cap radius r was evaluated within [5∘,30∘] [5 ,30 ] to assess its impact on coverage, accuracy, and efficiency. As shown in Fig. 1, performance forms a single peak near r=10∘r=10 . Both total accuracy and normalized alignment error remain stable between 9∘9 and 11∘11 , achieving the optimum at r=10∘r=10 (accT=81.48%acc_T=81.48\%, eADI/d=4.18×10−3e_ADI/d=4.18×10^-3). Smaller radii (≤8∘≤8 ) lead to redundant refinements without accuracy gain, while larger ones (≥12∘≥12 ) reduce coverage and precision. Runtime varies slightly (≈1.3−1.8s≈ 1.3-1.8s per model). The interval 9∘−11∘9 -11 thus provides the best trade-off among coverage, accuracy, and efficiency, and r=10∘r=10 is adopted as the default. 4.2.2 Order-Set Composition Table 1 summarizes the evaluation of different candidate order sets N, obtained by systematically reducing, expanding, and cross-validating their elements, including configurations with and without the 2-fold term. Reducing N to 3,4 \3,4 \ or 3,4,5 \3,4,5 \ caused clear accuracy drops, indicating that 3,4,5,6 \3,4,5,6 \ is the minimal effective configuration. Expanding the set to 3,4,5,6,7 \3,4,5,6,7 \ or 3,4,5,6,7,9 \3,4,5,6,7,9 \ increased runtime without measurable benefit, while 3,5,6 \3,5,6 \ further confirmed the necessity of the 4-fold term for distinguishing cubic and dihedral symmetries. Including n=2n=2 introduced instability, as trivial half-turn alignments dominated optimization and led to high-order shapes, particularly several 3-fold models, being misclassified as 2-fold. (Detailed analyses are provided in the supplementary material.) Overall, =3,4,5,6N= \3,4,5,6 \ provides the best balance among accuracy, stability, and computational efficiency, while adding 2-fold or rare high-order terms introduces misalignment or unnecessary cost. Table 2: Comprehensive evaluation of symmetry classification and axis localization across datasets. DSRSTOtex: texture subset. Method Dataset Category Statistics Dataset-Level Performance Name Object Type Count Errors accTacc_T acctacc_t accoacc_o accTacc_T time eADI/de_ADI/d (ours / Wang et al. [26]) BOP (original) BOP[27] C∞C_∞ 7 0 100.00% 80.00% 100.00% 80.00% — — D∞D_∞ 3 0 100.00% CnC_n 35 9 74.29% DnD_n 5 1 80.00% KASALv2(ours) BOP[27] C∞C_∞ 7 0 100.00% 92.00% 91.30% 84.00% 1.61 0.00256/0.00290 D∞D_∞ 3 0 100.00% CnC_n 35 8 77.14% DnD_n 5 0 100.00% KASALv2(ours) GSO[28] C∞C_∞ 118 3 97.46% 96.58% 98.10% 94.75% 1.37 0.00262/0.00256 D∞D_∞ 11 0 100.00% CnC_n 85 14 83.53% DnD_n 222 6 97.30% S4S_4 2 0 100.00% KASALv2(ours) DSRSTO[26] C∞C_∞ 1 0 100.00% 81.48% 100.00% 81.48% 1.46 0.00212/0.00172 D∞D_∞ 1 0 100.00% CnC_n 7 5 28.57% DnD_n 12 0 100.00% A4A_4 1 0 100.00% S4S_4 3 0 100.00% A5A_5 2 0 100.00% KASALv2(ours) DSRSTOtexDSRSTO_tex[26] DnD_n 11 3 72.73% 72.32% 87.50% 72.32% 1.16 0.00195/0.00181 4.3 Cross-Dataset Evaluation Cross-dataset evaluations were conducted on DSRSTO, BOP, and GSO. Table 2 reports the metrics defined in Sec. 4.1.3, the average runtime per object, and localization accuracy compared with the reference method [26]. On DSRSTO, the method achieved 81.48% accuracy with eADI/d=0.00212e_ADI/d=0.00212. On BOP, under our stricter re-annotation of 50 symmetric instances, it reached 84.00% accuracy, improving over 80.00% under the original labels and reducing eADI/de_ADI/d from 0.00290 to 0.00256. On GSO, it obtained 94.75% accuracy with eADI/d=0.00262e_ADI/d=0.00262. The 72.32% accuracy on DSRSTOtex further confirms applicability to texture-level symmetries. The lower DSRSTO performance mainly stems from challenging high-order and visually ambiguous CnC_n cases; consistently, CnC_n is also the lowest-accuracy category in BOP. Details of the original BOP annotations, our re-annotation, and the full error analysis are provided in the supplementary material. 4.4 Effect of Symmetry Priors on Pose Estimation To assess the downstream value of automatically estimated symmetry priors, we train two identical FoundationPose [29] models on a synthetic GSO training set of about one million rendered images, differing only in the use of KASALv2-derived rotational-symmetry priors. The prior is injected only at the label level via ZebraPose’s symmetry-aware label canonicalization module, which maps each GT pose to a canonical representative in its symmetry equivalence class for both discrete and continuous symmetries. All other settings remain unchanged. Table 3 reports a retrained multi-checkpoint evaluation on five BOP datasets. Averaged over five checkpoints from the same training run, the results improve on all five datasets, with the largest mean gain of 0.9 points on IC-BIN. Table 3: FoundationPose with and without KASALv2 symmetry priors on five BOP datasets. Results are reported as mean (range) over five checkpoints from the same training run (seed 0). Dataset w/ KASALv2 mean (range) w/o KASALv2 mean (range) Diff. LM-O 71.9 (71.7–72.0) 71.6 (71.5–71.8) +0.3 IC-BIN 64.7 (64.4–65.3) 63.8 (63.2–64.4) +0.9 YCB-V 87.7 (87.6–87.9) 87.6 (87.5–87.8) +0.1 ITODD 65.8 (65.1–66.4) 65.0 (64.9–65.2) +0.8 T-LESS 59.3 (58.6–60.1) 58.8 (58.3–59.3) +0.5 5 Conclusion This work presents a unified and fully automatic framework for 3D rotational symmetry analysis, jointly addressing symmetry-type classification, rotational-order identification, and axis localization across all eight canonical types. The Geometric Degeneration Hierarchy (GDH), together with an integrated inference pipeline, enables complete and reference-free symmetry reasoning for real-world 3D objects. Extensive experiments show strong accuracy and generalization, reaching 94.75% on GSO, and the symmetry priors produced by KASALv2 improve FoundationPose accuracy by up to 0.9% across five BOP datasets. Despite these results, characteristic failure modes remain, including the underestimation of high rotational orders and sensitivity to local geometric irregularities. Future work will explore more frequency-sensitive order estimation and improved robustness to locally perturbed structures. 6 Acknowledgments This work was supported by the National Natural Science Foundation of China under Grant No. 52375487 and the Special Fund of Jiangsu Province for Key Research and Development under Grant No. BE2023041. The authors also thank the reviewers for their thoughtful and constructive comments, which have helped improve this paper. Due to space limitations, several complementary analyses and extended discussions could not be included in the present conference version and will be reported in a future journal extension, including posterior verification for continuous symmetries and secondary two-fold axes, robustness under geometric perturbations, additional optimizer and runtime analyses, large-scale rotational-symmetry annotation on Objaverse, and broader downstream evaluation studies. References [1] T. Bertens, B. Caasenbrood, A. Saccon, and A. Jalba (2025) Symmetry-induced ambiguity in orientation estimation from RGB images. Mach. Vis. Appl. 36, p. 40. Cited by: §1, §2.3. [2] W. Kehl, F. Manhardt, F. Tombari, S. Ilic, and N. Navab (2017) SSD-6D: making RGB-based 3D detection and 6D pose estimation great again. In ICCV, p. 1530–1538. Cited by: §1, §2.3. [3] M. Rad and V. Lepetit (2017) B8: a scalable, accurate, robust to partial occlusion method for predicting the 3D poses of challenging objects without using depth. In ICCV, p. 3828–3836. Cited by: §1, §2.3. [4] L. Mescheder, M. Oechsle, M. Niemeyer, S. Nowozin, and A. Geiger (2019) Occupancy networks: learning 3D reconstruction in function space. In CVPR, p. 4460–4470. Cited by: §1, §2.3. [5] F. Manhardt, D. M. Arroyo, C. Rupprecht, B. Busam, T. Birdal, N. Navab, and F. Tombari (2019) Explaining the ambiguity of object detection and 6D pose from visual data. In ICCV, p. 6840–6849. Cited by: §1. [6] T. Hsiao, H. Chen, H. Yang, and C. Lee (2024) Confronting ambiguity in 6D object pose estimation via score-based diffusion on SE(3). In CVPR, p. 352–362. Cited by: §1. [7] G. Pitteri, M. Ramamonjisoa, S. Ilic, and V. Lepetit (2019) On object symmetries and 6D pose estimation from images. In 3DV, p. 614–622. Cited by: §1, §2.3. [8] Y. Xiang, T. Schmidt, V. Narayanan, and D. Fox (2018) PoseCNN: a convolutional neural network for 6D object pose estimation in cluttered scenes. In Proceedings of Robotics: Science and Systems, Cited by: §1, §2.3, §4.1.3. [9] T. Hodan, M. Sundermeyer, B. Drost, Y. Labbé, E. Brachmann, F. Michel, C. Rother, and J. Matas (2020) BOP challenge 2020 on 6D object localization. In Computer Vision – ECCV 2020 Workshops, p. 577–594. Cited by: §1, §2.3. [10] Y. Wang, M. Hu, H. Li, and C. Luo (2025) HccePose(BF): predicting front & back surfaces to construct ultra-dense 2D-3D correspondences for pose estimation. In ICCV, p. 7166–7175. Cited by: §1, §2.3. [11] Y. Su, M. Saleh, T. Fetzer, J. Rambach, N. Navab, B. Busam, D. Stricker, and F. Tombari (2022) ZebraPose: coarse to fine surface encoding for 6DoF object pose estimation. In CVPR, p. 6728–6738. Cited by: §1, §2.3. [12] J. E. Humphreys (1990) Reflection groups and coxeter groups. Cambridge University Press. Cited by: §1, §3.1. [13] H. S. M. Coxeter (1973) Regular polytopes. Dover Publications. Cited by: §1, §3.1. [14] Jr. Miller (1972) Symmetry groups and their applications. Academic Press. Cited by: §1, §3.1. [15] D. Montgomery and C. T. Yang (1962) A theorem on the action of SO(3). Pacific J. Math. 12 (4), p. 1385–1400. Cited by: §1, §3.1. [16] H. Weyl (1952) Symmetry. Princeton University Press. Cited by: §1, §3.1. [17] R. Gothandaraman, R. Jha, and S. Muthuswamy (2020) Reflectional and rotational symmetry detection of cad models based on point cloud processing. In IEEE 4th Conf. Inf. Commun. Technol., p. 1–5. Cited by: §1, §2.1, §2. [18] L. Hruda, I. Kolingerová, M. Lávička, and M. Maňák (2022) Rotational symmetry detection in 3D using reflectional symmetry candidates and quaternion-based rotation parameterization. Comput. Aided Geom. Des. 98, p. 102138. Cited by: §1, §2.1, §2. [19] A. Ecins, C. Fermüller, and Y. Aloimonos (2018) Seeing behind the scene: using symmetry to reason about objects in cluttered environments. In IEEE/RSJ Int. Conf. Intell. Robots Syst., p. 7193–7200. Cited by: §1, §2.2, §2. [20] I. Sipiran (2018) Completion of cultural heritage objects with rotational symmetry. In Eurographics Workshop on 3D Object Retrieval, p. 87–93. Cited by: §1, §2.2, §2. [21] I. Sipiran (2018) Analysis of partial axial symmetry on 3D surfaces and its application in the restoration of cultural heritage objects. In IEEE Int. Conf. Comput. Vis. Worksh., p. 2925–2933. Cited by: §1, §2.2, §2. [22] S. Mansouri and H. Ebrahimnezhad (2016) Efficient axial symmetry aware mesh approximation with application to 3D pottery models. Multimed. Tools Appl. 75 (14), p. 8347–8379. Cited by: §1, §2.2, §2. [23] Y. Shi, J. Huang, H. Zhang, X. Xu, S. Rusinkiewicz, and K. Xu (2020) SymmetryNet: learning to predict reflectional and rotational symmetries of 3D shapes from single-view RGB-D images. ACM TOG 39 (6), p. 213:1–213:14. Cited by: §1, §2.1, §2. [24] A. Martinet, C. Soler, N. Holzschuch, and F. X. Sillion (2006) Accurate detection of symmetries in 3D shapes. ACM TOG 25 (2), p. 439–464. Cited by: §1, §2.1, §2. [25] S. Korman, R. Litman, S. Avidan, and A. Bronstein (2015) Probably approximately symmetric: fast rigid symmetry detection with global guarantees. Comput. Graph. Forum 34 (1), p. 2–13. Cited by: §1, §2.1, §2. [26] Y. Wang and C. Luo (2024) Key-axis-based localization of symmetry axes in 3D objects utilizing geometry and texture. IEEE TIP 33, p. 6720–6733. Cited by: §1, §1, §1, §2.2, §2, §4.1.3, §4.3, Table 2, Table 2, Table 2, §4. [27] T. Hodaň, F. Michel, E. Brachmann, W. Kehl, A. G. Buch, D. Kraft, B. Drost, J. Vidal, S. Ihrke, X. Zabulis, C. Sahin, F. Manhardt, F. Tombari, T. Kim, J. Matas, and C. Rother (2018) BOP: benchmark for 6D object pose estimation. In ECCV, p. 19–35. Cited by: §1, Table 2, Table 2, §4. [28] L. Downs, A. Francis, N. Koenig, B. Kinman, R. Hickman, K. Reymann, T. B. McHugh, and V. Vanhoucke (2022) Google scanned objects: a high-quality dataset of 3D scanned household items. In IEEE Int. Conf. Robot. Autom., p. 2553–2560. Cited by: §1, §4.1.3, Table 2, §4. [29] B. Wen, W. Yang, J. Kautz, and S. Birchfield (2024) FoundationPose: unified 6D pose estimation and tracking of novel objects. In CVPR, p. 17868–17879. Cited by: §1, §4.4. [30] S. M. Hosseini, S. Nasiri, R. Hosseini, and H. Moradi (2025) Single-view 3D reconstruction of surface of revolution. Vis. Comput. 41, p. 8961–8973. Cited by: §2.2. [31] G. Wang, F. Manhardt, F. Tombari, and X. Ji (2021) GDR-Net: geometry-guided direct regression network for monocular 6D object pose estimation. In CVPR, p. 16606–16616. Cited by: §2.3. [32] T. Hodaň, D. Baráth, and J. Matas (2020) EPOS: estimating 6D pose of objects with symmetries. In CVPR, p. 11700–11709. Cited by: §2.3. [33] N. Mo, W. Gan, N. Yokoya, and S. Chen (2022) ES6D: a computation efficient and symmetry-aware 6D pose regression framework. In CVPR, p. 6718–6727. Cited by: §2.3. [34] J. Richter-Klug and U. Frese (2021) Handling object symmetries in CNN-based pose estimation. In IEEE Int. Conf. Robot. Autom., p. 13850–13856. Cited by: §2.3. [35] F. Klein (1956) Lectures on the icosahedron and the solution of equations of the fifth degree. Dover Publications. Cited by: §3.1. [36] L. Schläfli (1901) Theorie der vielfachen kontinuität. Zürcher & Furrer. Cited by: §3.1. [37] H. F. Blichfeldt (1917) Finite collineation groups: with an introduction to the theory of groups of operators and substitution groups. The University of Chicago Press. Cited by: §3.1. [38] J. H. Conway and N. J. A. Sloane (1999) Sphere packings, lattices and groups. Springer. Cited by: §3.1. [39] H. G. Barrow, J. M. Tenenbaum, R. C. Bolles, and H. C. Wolf (1977) Parametric correspondence and chamfer matching: two new techniques for image matching. In IJCAI, p. 659–663. Cited by: §3.2.1. [40] B. Drost, M. Ulrich, P. Bergmann, P. Härtinger, and C. Steger (2017) Introducing MVTec ITODD—a dataset for 3D object recognition in industry. In IEEE Int. Conf. Comput. Vis. Worksh., p. 2200–2208. Cited by: §4.1.3. [41] T. Hodaň, P. Haluza, Š. Obdržálek, J. Matas, M. Lourakis, and X. Zabulis (2017) T-LESS: an RGB-D dataset for 6D pose estimation of texture-less objects. In IEEE Winter Conf. Appl. Comput. Vis., p. 880–888. Cited by: §4.1.3. [42] E. Brachmann, A. Krull, F. Michel, S. Gumhold, J. Shotton, and C. Rother (2014) Learning 6D object pose estimation using 3D object coordinates. In ECCV, p. 536–551. Cited by: §4.1.3. [43] R. Kaskman, S. Zakharov, I. Shugurov, and S. Ilic (2019) HomebrewedDB: RGB-D dataset for 6D pose estimation of 3D objects. In IEEE/CVF Int. Conf. Comput. Vis. Worksh., p. 2767–2776. Cited by: §4.1.3. [44] A. Doumanoglou, R. Kouskouridas, S. Malassiotis, and T. Kim (2016) Recovering 6D object pose and predicting next-best-view in the crowd. In CVPR, p. 3583–3592. Cited by: §4.1.3.