Research map / Point clouds / Shape and surface completion
Sketch and Text Guided Diffusion Model for Colored Point Cloud Generation
Zijie Wu, Yaonan Wang, Mingtao Feng, He Xie, Ajmal Mian
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 capsule attention routing on top of ResNet50 sketch encoder (ablation) | STPD sketch encoder · ShapeNet chair+table | CD (x10^3): 6.409 -> 1.935 | Improved |
| use proposed attention fusion instead of concatenating sketch and text embeddings (ablation) | STPD conditioning · ShapeNet chair+table | CD (x10^3): 3.748 -> 1.390 | Improved |
| add sketch conditioning to text-only conditioning (ablation) | STPD · ShapeNet chair+table | CD (x10^3): 28.325 -> 10.965 | Improved |
| add BERT text to simple C1x1 sketch encoder (ablation) | STPD with C1x1 sketch encoder · ShapeNet chair+table | CD (x10^3): 10.965 -> 9.651 | Improved |
| add text to attention-routing sketch encoder with fusion (ablation) | STPD · ShapeNet chair+table | CD (x10^3): 1.689 -> 1.390 | Improved |
| replace ResNet50 with simple C1x1 encoder when using attention routing (ablation) | STPD sketch encoder · ShapeNet chair+table | CD (x10^3): 1.935 -> 1.689 (C1x1 no text) | Improved |
| joint 6D diffusion of shape and color instead of staged diffusion (ablation) | STPD · ShapeNet chair+table | point cloud detail | Hurt |
| extend DPM to 6D color without staging | DPM · ShapeNet chair+table | MMD-CD (x10^3): 3.764 -> 4.765 | Hurt |
| capsule attention increases instance attention on sparse sketches | STPD · sketch attention | instance score (%): 1.07% -> 58.68% | Improved |
Closest papers by meaning
- TIGER: Time-Varying Denoising Model for 3D Point Cloud Generation with Diffusion Process (CVPR 2024 · similarity 0.88)
- Generating Images with 3D Annotations Using Diffusion Models (ICLR 2024 · similarity 0.87)
- PI3D: Efficient Text-to-3D Generation with Pseudo-Image Diffusion (CVPR 2024 · similarity 0.87)
- Point Cloud Pre-training with Diffusion Models (CVPR 2024 · similarity 0.87)
- Neural Point Cloud Diffusion for Disentangled 3D Shape and Appearance Generation (CVPR 2024 · similarity 0.87)
- SuperPC: A Single Diffusion Model for Point Cloud Completion, Upsampling, Denoising, and Colorization (CVPR 2025 · similarity 0.87)
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.
