Research map / Architectures and efficiency / CNN design
Axis-level Symmetry Detection with Group-Equivariant Representation
Wongyun Yu, Ahyun Seo, Minsu Cho
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add orientational anchors to axis-level reflection detection (ablation) | Axis-level detection baseline · DENDI reflection | sAP@5/10/15: 6.2 -> 16.6, 9.3 -> 19.9, 11.2 -> 21.1 | Improved |
| add single-kernel reflectional matching (k=0) (ablation) | Axis-level detection + orientational anchors · DENDI reflection | sAP@5/10/15: 16.6 -> 17.6, 19.9 -> 20.7, 21.1 -> 21.8 | Improved |
| expand reflectional matching to kernels k=0,1 (ablation) | Reflectional matching k=0 · DENDI reflection | sAP@5/10/15: 17.6 -> 18.4, 20.7 -> 22.0, 21.8 -> 23.7 | Improved |
| expand reflectional matching to kernels k=0,1,2 (ablation) | Reflectional matching k=0,1 · DENDI reflection | sAP@5/10/15: 18.4 -> 18.8, 22.0 -> 22.7, 23.7 -> 24.7 | Improved |
| add single-kernel rotational matching (k=0) (ablation) | Axis-level detection baseline · DENDI rotation center | sAP@5/10/15: 31.5 -> 35.9, 34.7 -> 37.8, 35.7 -> 37.0 | Improved |
| expand rotational matching to kernels k=0,1 (ablation) | Rotational matching k=0 · DENDI rotation center | sAP@5/10/15: 35.9 -> 36.2, 37.8 -> 38.2, 37.0 -> 37.4 | Improved |
| expand rotational matching to kernels k=0,1,2 (ablation) | Rotational matching k=0,1 · DENDI rotation center | sAP@5/10/15: 36.2 -> 36.8, 38.2 -> 39.1, 37.4 -> 40.0 | Improved |
| full rotational matching (k=0,1,2) on fold classification (ablation) | Axis-level detection baseline · DENDI rotation fold | fold sAP@5/10/15: 22.5 -> 26.6, 24.6 -> 28.3, 25.3 -> 28.9 | Improved |
| use group-equivariant design instead of non-equivariant variant (ablation) | Proposed model · DENDI test rotation robustness | RMSE of length map: significantly lower RMSE (no numbers given) | Improved |
| axis-level detection instead of heatmap-based PMCNet | Ours vs PMCNet · LDRS reflection | F1: 37.3 -> 43.4 | Improved |
| axis-level detection instead of heatmap-based PMCNet | Ours vs PMCNet · SDRW reflection | F1: 68.8 -> 68.3 | No clear effect |
| axis-level detection instead of EquiSym | Ours vs EquiSym · DENDI rotation | F1: 22.4 -> 26.8 | Improved |
Closest papers by meaning
- R2Det: Exploring Relaxed Rotation Equivariance in 2D Object Detection (ICLR 2025 · similarity 0.89)
- Leveraging 3D Geometric Priors in 2D Rotation Symmetry Detection (CVPR 2025 · similarity 0.89)
- Relaxed Rotational Equivariance via G-Biases in Vision (AAAI 2025 · similarity 0.88)
- Symmetry Strikes Back: From Single-Image Symmetry Detection to 3D Generation (CVPR 2025 · similarity 0.88)
- Symmetry Discovery Beyond Affine Transformations (NeurIPS 2024 · similarity 0.86)
- AtlasD: Automatic Local Symmetry Discovery (ICML 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
