Research map / Point clouds / Shape and surface completion
UNIST: Unpaired Neural Implicit Shape Translation Network
Qimin Chen, Johannes Merz, Aditya Sanghi, Hooman Shayani, Ali Mahdavi-Amiri, Hao Zhang
Check 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 position-aware latent grids instead of regular holistic encoding (ablation) | UNIST shape translation · 2D font translation: Solid ↔ Dotted, Regular ↔ Italic, Regular ↔ Bold, Sharp ↔ Smooth | preservation of local geometric features | Improved |
| Use position-aware latent grids instead of regular holistic encoding (ablation) | UNIST shape translation · 3D Chair ↔ Table, Armrest, and Tall table ↔ Short table | preservation of geometric and structural features | Improved |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · A ↔ H | MSE: 0.184 -> 0.215 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · A ↔ H | IoU: 0.507 -> 0.441 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · G ↔ R | MSE: 0.197 -> 0.211 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · G ↔ R | IoU: 0.493 -> 0.452 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · M ↔ N | MSE: 0.201 -> 0.233 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · M ↔ N | IoU: 0.508 -> 0.433 | Hurt |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · Chair with/without Armrest | one-sided CD: 0.0255 -> 0.0234; 0.0267 -> 0.0235 | Improved |
| Use position-aware encoding instead of regular encoding (ablation) | UNIST · Chair ↔ Table | user preference: ranked 1st votes: 31.34% -> 38.11%; 27.78% -> 36.28% | Improved |
| Represent shapes as neural implicit fields rather than point clouds | UNIST versus LOGAN · 2D shape translation | visual translation quality and fine/topological feature reproduction | Improved |
| Translate on latent grids rather than holistic latent vectors (ablation) | UNIST · unpaired shape translation | spatial feature and local-detail preservation | Improved |
Closest papers by meaning
- Improving Shape Deformation in Unsupervised Image-to-Image Translation (ECCV 2018 · similarity 0.87)
- Cross-Domain Cascaded Deep Translation (ECCV 2020 · similarity 0.86)
- RIM-Net: Recursive Implicit Fields for Unsupervised Learning of Hierarchical Shape Structures (CVPR 2022 · similarity 0.86)
- Semantic-Aware Implicit Template Learning via Part Deformation Consistency (ICCV 2023 · similarity 0.86)
- Deformed Implicit Field: Modeling 3D Shapes With Learned Dense Correspondence (CVPR 2021 · similarity 0.86)
- Domain-Scalable Unpaired Image Translation via Latent Space Anchoring (TPAMI 2023 · 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.
