Research map / 3D vision / Face and shape
GarNet++: Improving Fast and Accurate Static 3D Cloth Draping by Curvature Loss
Erhan Gündoğdu, Victor Constantin, Shaifali Parashar, Amrollah Seifoddini, Minh Dang, Mathieu Salzmann, Pascal Fua
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 mesh convolution patch-wise garment features (Late -> Global) (ablation) | GarNet · Own dataset (SMPL), Jeans | E_dist (cm): 0.70 -> 0.42 | Improved |
| add explicit nearest-neighbour local body feature pooling (Global -> Local) (ablation) | GarNet · Own dataset (SMPL), Sweater | E_dist (cm): 0.93 -> 0.85 | Improved |
| use SMPL parameters as body input instead of point features (ablation) | GarNet-Local · Own dataset (SMPL), Jeans | E_dist (cm): 0.41 -> 0.44 | No clear effect |
| add normal and bending terms to vert+pen loss (ablation) | GarNet-Local · Wang dataset | E_norm (deg): 10.01 -> 7.34 | Improved |
| add normal term to vert+pen+bend loss (ablation) | GarNet-Local · Wang dataset | E_norm (deg): 9.06 -> 7.34 | Improved |
| add bending term to vert+pen+norm loss (ablation) | GarNet-Local · Wang dataset | E_norm (deg): 7.91 -> 7.34 | Improved |
| add penetration term to vert+norm+bend loss (ablation) | GarNet-Local · Wang dataset | E_dist: 0.48 -> 0.42 | Improved |
| add RQ curvature loss to full loss (ablation) | GarNet-Local · Wang dataset | E_dist and E_norm: 0.42 -> 0.45 E_dist; 7.34 -> 7.2 E_norm | Mixed |
| use RQ losses at multiple neighbourhood sizes (8+16+32) instead of 8-NN only (ablation) | GarNet-Local-RQ · Own dataset (SMPL) | E_norm (deg): 7.51 -> 7.2 | Improved |
| add RQ curvature fine-tuning | GarNet-Local · Own dataset (SMPL), T-shirt | E_norm (deg): 8.34 -> 7.79 E_norm; 0.56 -> 0.63 E_dist | Mixed |
| use RQ loss instead of mean curvature normal loss (ablation) | GarNet-Local · T-shirt (SMPL) | E_norm (deg): 7.66 -> 7.79 E_norm; 0.26 -> 0.09 L_RQ | Mixed |
| use RQ loss instead of uniform mean curvature loss (ablation) | GarNet-Local · T-shirt (SMPL) | E_norm (deg): 8.25 -> 7.79 | Improved |
Closest papers by meaning
- GarNet: A Two-Stream Network for Fast and Accurate 3D Cloth Draping (ICCV 2019 · similarity 0.95)
- ISP: Multi-Layered Garment Draping with Implicit Sewing Patterns (NeurIPS 2023 · similarity 0.90)
- GarmentCodeData: A Dataset of 3D Made-to-Measure Garments With Sewing Patterns (ECCV 2024 · similarity 0.89)
- CLOTH4D: A Dataset for Clothed Human Reconstruction (CVPR 2023 · similarity 0.89)
- DeepWrinkles: Accurate and Realistic Clothing Modeling (ECCV 2018 · similarity 0.89)
- DeClotH: Decomposable 3D Cloth and Human Body Reconstruction from a Single Image (CVPR 2025 · similarity 0.88)
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.
