Research map / 3D vision / Face and shape
Cross-Modal 3D Shape Generation and Manipulation
Zezhou Cheng, Menglei Chai, Jian Ren, Hsin-Ying Lee, Kyle Olszewski, Zeng Huang, Subhransu Maji, Sergey Tulyakov
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 multi-trial latent optimization (pick lowest-loss code) (ablation) | MM-VADs single-view sketch reconstruction · ShapeNet airplanes | mean Chamfer distance: 5.50 -> 1.73 | Improved |
| add multi-trial latent optimization (pick lowest-loss code) (ablation) | MM-VADs single-view sketch reconstruction · ShapeNet chairs | mean Chamfer distance: 9.10 -> 4.70 | Improved |
| add latent regularization loss gamma*max(||z||^2,beta) (ablation) | MM-VADs single-view reconstruction · ShapeNet single-view reconstruction | reconstruction success | Improved |
| use 128-dim instead of 32-dim latent codes (ablation) | MM-VADs · ShapeNet shape reconstruction | shape reconstruction quality | Improved |
| replace task-specific encoder-decoder with MM-VADs latent optimization under 1/2-horizontal occlusion | single-view reconstruction · ShapeNet, 1/2-horizontal input | avg Chamfer distance x1e3: 8.36 -> 3.56 | Improved |
| replace encoder-decoder with MM-VADs under 1/4-vertical occlusion | single-view reconstruction · ShapeNet, 1/4-vertical input | avg Chamfer distance x1e3: 11.61 -> 5.60 | Improved |
| replace encoder-decoder with MM-VADs on full-view input | single-view reconstruction · ShapeNet, Full input | avg Chamfer distance x1e3: 2.59 -> 3.44 | Hurt |
| apply few-shot adaptation with 10 2D images (MineGAN-style mapping) | MM-VADs · Few-shot generation (avg) | classification error: 0.75 -> 0.04 | Improved |
| apply few-shot adaptation with 10 2D images | MM-VADs · Few-shot generation (avg) | FID: 109.0 -> 105.3 | Improved |
| edit color with scribbles via latent optimization | MM-VADs · Color editing, airplane | PSNR: 19.84 -> 26.41 | Improved |
| edit color with scribbles via latent optimization | MM-VADs · Color editing, chair | LPIPS: 0.33 -> 0.20 | Improved |
| sketch edit removing airplane engine | MM-VADs · Sketch shape editing, airplane - engine | Chamfer distance to manual edit: 0.096 -> 0.059 | Improved |
Closest papers by meaning
- Doodle Your 3D: From Abstract Freehand Sketches to Precise 3D Shapes (CVPR 2024 · similarity 0.89)
- Interactive3D: Create What You Want by Interactive 3D Generation (CVPR 2024 · similarity 0.86)
- Image Sculpting: Precise Object Editing with 3D Geometry Control (CVPR 2024 · similarity 0.86)
- ShapeShifter: 3D Variations Using Multiscale and Sparse Point-Voxel Diffusion (CVPR 2025 · similarity 0.86)
- Instant3dit: Multiview Inpainting for Fast Editing of 3D Objects (CVPR 2025 · similarity 0.85)
- ShapeCraft: LLM Agents for Structured, Textured and Interactive 3D Modeling (NeurIPS 2025 · similarity 0.85)
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.
