Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-supervised Garment Dynamics with Persistent Wrinkles
Xiaoyuan Yang, Deshan Gong, Taku Komura, He Wang
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 |
|---|---|---|---|
| use hinge-graph GNN message passing instead of RNN for P-Net (ablation) | P-Net · Unseen motions | BED (rad): 0.86971 -> 0.28714 | Improved |
| use hinge-graph GNN message passing instead of LSTM for P-Net (ablation) | P-Net · Unseen motions | BED (rad): 0.84272 -> 0.28714 | Improved |
| use hinge-graph GNN instead of Transformer for P-Net (ablation) | P-Net · Unseen motions | BED (rad): 0.52095 -> 0.28714 | Improved |
| use hinge-graph GNN instead of Transformer for P-Net (ablation) | P-Net · Unseen motions | FPS: 0.17 -> 9.43 | Improved |
| use hinge-graph GNN instead of Transformer for P-Net (ablation) | P-Net · Unseen motions | GPU memory (GB): 9.2 -> 0.6 | Improved |
| use hinge-graph GNN instead of RNN for P-Net (ablation) | P-Net · Unseen motions | BE (rad): 0.17048 -> 0.17118 | Hurt |
| increase message-passing steps from 1 to 3 (ablation) | P-Net · P-Net RB prediction | mean RB error (rad): 0.0281575 -> 0.017319 | Improved |
| increase message-passing steps from 2 to 3 (ablation) | P-Net · P-Net RB prediction | max RB error (rad): 3.0522114 -> 2.975552 | Improved |
| increase alternating training cycles beyond four (ablation) | EPNet training · Convergence | training convergence | No clear effect |
| add curriculum learning from elasticity to elasto-plasticity (ablation) | EPNet training · Training stability | convergence stability | Improved |
| model plasticity with learned rest bending versus best elastic baseline | EPNet · Unseen motions | BED (rad): 45.34% in BED, 74.36% in PE, 2.79% in BE, 26.19% in MED, 32.50% in CD, 12.10% in BEN | Improved |
| model plasticity with learned rest bending versus best elastic baseline | EPNet · Unseen bodies | BED (rad): 29.48% in BED, 32.78% in PE, 0.82% in BE, 13.38% in MED, 25.00% in CD, 18.93% in BEN | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SENC — SENC: Handling Self-collision in Neural Cloth Simulation (ECCV 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- SNUG: Self-Supervised Neural Dynamic Garments (CVPR 2022 · similarity 0.92)
- NeuralGarSim: Geometry-agnostic Garment Simulation with Neural Fields (ECCV 2026 · similarity 0.90)
- Learning 3D Garment Animation from Trajectories of A Piece of Cloth (NeurIPS 2024 · similarity 0.90)
- NeuralClothSim: Neural Deformation Fields Meet the Thin Shell Theory (NeurIPS 2024 · similarity 0.88)
- GarmentLab: A Unified Simulation and Benchmark for Garment Manipulation (NeurIPS 2024 · similarity 0.87)
- DexGarmentLab: Dexterous Garment Manipulation Environment with Generalizable Policy (NeurIPS 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.
