Research map / 3D vision / Face and shape
PASTA: Part-Aware Sketch-to-3D Shape Generation with Text-Aligned Prior
Seunggwan Lee, Hwanhee Jung, Byoungsoo Koh, Qixing Huang, Sang Ho Yoon, Sangpil Kim
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 IndivGCN (ablation) | PASTA baseline · AmateurSketch-3D | FID: 169.2 -> 158.8 | Improved |
| add IndivGCN (ablation) | PASTA baseline · ProSketch-3D | FID: 157.9 -> 148.4 | Improved |
| add PartGCN (ablation) | PASTA baseline · AmateurSketch-3D | FID: 169.2 -> 163.1 | Improved |
| add PartGCN (ablation) | PASTA baseline · ProSketch-3D | FID: 157.9 -> 152.0 | Improved |
| combine IndivGCN and PartGCN as ISG-Net (ablation) | PASTA baseline · ProSketch-3D | FID: 148.4 -> 129.2 | Improved |
| add text-aligned prior from VLM (ablation) | PASTA baseline · ProSketch-3D | FID: 157.9 -> 135.5 | Improved |
| add text-aligned prior on top of ISG-Net (full PASTA) (ablation) | PASTA baseline · ProSketch-3D | FID: 129.2 -> 112.2 | Improved |
| add ISG-Net on top of text prior (full PASTA) (ablation) | PASTA baseline · AmateurSketch-3D | CD: 0.101 -> 0.090 | Improved |
| sequential instead of parallel text-visual cross-attention (ablation) | PASTA baseline · ProSketch-3D | FID: 120.7 -> 112.2 | Improved |
| use LLaVA-7B instead of CLIP as VLM (ablation) | PASTA baseline · AmateurSketch-3D | FID: 152.1 -> 143.9 | Improved |
| use LLaVA-7B instead of BLIP2 T5 (ablation) | PASTA baseline · ProSketch-3D | FID: 115.6 -> 112.2 | Improved |
| use LLaVA-7B instead of LLaVA-13B (ablation) | PASTA baseline · AmateurSketch-3D | FID: 144.3 -> 143.9 | No clear effect |
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.
- Sketch2Mesh — Sketch2Mesh: Reconstructing and Editing 3D Shapes From Sketches (ICCV 2021) · better on 6 of 6 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
- Michelangelo: Conditional 3D Shape Generation based on Shape-Image-Text Aligned Latent Representation (NeurIPS 2023 · similarity 0.90)
- ShapeScaffolder: Structure-Aware 3D Shape Generation from Text (ICCV 2023 · similarity 0.89)
- ShapeWords: Guiding Text-to-Image Synthesis with 3D Shape-Aware Prompts (CVPR 2025 · similarity 0.87)
- ShapeWalk: Compositional Shape Editing Through Language-Guided Chains (CVPR 2024 · similarity 0.86)
- ArtFormer: Controllable Generation of Diverse 3D Articulated Objects (CVPR 2025 · similarity 0.86)
- PartGen: Part-level 3D Generation and Reconstruction with Multi-view Diffusion Models (CVPR 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.
