Research map / Detection and segmentation / Open-set and few-shot detection
Learning Single/Multi-Attribute of Object With Symmetry and Group
Yong–Lu Li, Yue Xu, Xinyu Xu, Xiaohan Mao, Cewu Lu
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 symmetry loss L_sym (ablation) | SymNet · aPY | mAUC (%): 65.7 -> 86.1 | Improved |
| add symmetry loss L_sym (ablation) | SymNet · MIT-States CZSL | top-1 accuracy (%): 18.3 -> 19.9 | Improved |
| add group axiom losses (ablation) | SymNet · MIT-States CZSL | top-1 accuracy (%): 16.9 -> 19.9 | Improved |
| add classification loss L_cls (ablation) | SymNet · MIT-States CZSL | top-1 accuracy (%): 10.3 -> 19.9 | Improved |
| add classification loss L_cls (ablation) | SymNet · UT-Zappos CZSL | top-1 accuracy (%): 28.7 -> 52.1 | Improved |
| add multi-attribute correlation losses L_tri^sym and L_tri^corr (ablation) | SymNet · aPY | mAUC (%): 82.1 -> 86.1 | Improved |
| add multi-attribute correlation losses L_tri^sym and L_tri^corr (ablation) | SymNet · SUN | mAUC (%): 88.1 -> 88.4 | Improved |
| add attribute-as-attention (ablation) | SymNet · UT-Zappos CZSL | top-1 accuracy (%): 48.5 -> 52.1 | Improved |
| use tanh instead of sigmoid attention (ablation) | SymNet · UT-Zappos CZSL | top-1 accuracy (%): 52.1 -> 42.0 | Hurt |
| use L1 distance instead of L2 (ablation) | SymNet · MIT-States CZSL | top-1 accuracy (%): 19.9 -> 7.1 | Hurt |
| use cosine distance instead of L2 (ablation) | SymNet · UT-Zappos CZSL | top-1 accuracy (%): 52.1 -> 18.7 | Hurt |
| replace SymNet losses with AttrOperator losses (ablation) | SymNet · MIT-States CZSL | top-1 accuracy (%): 19.9 -> 17.0 | Hurt |
Closest papers by meaning
- Symmetry and Group in Attribute-Object Compositions (CVPR 2020 · similarity 0.98)
- MatrixNet: Learning over symmetry groups using learned group representations (NeurIPS 2024 · similarity 0.85)
- A Generative Model of Symmetry Transformations (NeurIPS 2024 · similarity 0.84)
- Learning Infinitesimal Generators of Continuous Symmetries from Data (NeurIPS 2024 · similarity 0.83)
- Disentangled Representation Learning via Modular Compositional Bias (NeurIPS 2025 · similarity 0.83)
- Learning symmetries via weight-sharing with doubly stochastic tensors (NeurIPS 2024 · similarity 0.83)
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.
