Research map / 3D vision / Face and shape
Towards Implicit Text-Guided 3D Shape Generation
Zhengzhe Liu, Yi Wang, Xiaojuan Qi, Chi-Wing Fu
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 |
|---|---|---|---|
| Add joint training with a pretrained auto-encoder (ablation) | Text-guided shape generation · Text-guided shape generation | IoU: 0.03 -> 12.04 | Improved |
| Add decoupled shape-color decoders (ablation) | Text-guided shape generation · Text-guided shape generation | PS: 2.95$\pm$0.03 -> 3.16$\pm$0.04 | Improved |
| Add decoupled shape-color decoders (ablation) | Text-guided shape generation · Text-guided shape generation | FPD: 35.05 -> 31.09 | Improved |
| Add the word-level spatial transformer (ablation) | Text-guided shape generation · Text-guided shape generation | FPD: 31.09 -> 30.34 | Improved |
| Add cyclic loss (ablation) | Text-guided shape generation · Text-guided shape generation | IoU: 12.24 -> 12.33 | Improved |
| Add cyclic loss (ablation) | Text-guided shape generation · Text-guided shape generation | FPD: 30.34 -> 30.80 | Hurt |
| Replace FC IMLE with style-based shape IMLE (ablation) | Diversified shape generation · Diversified shape generation | PS: 2.93 $\pm$ 0.02 -> 3.39 $\pm$ 0.02 | Improved |
| Replace Latent GAN with shape IMLE (ablation) | Diversified shape generation · Diversified shape generation | R-Precision: 21.20 $\pm$ 0.11 -> 27.60 $\pm$ 0.39 | Improved |
| Add the word-level spatial transformer (ablation) | Diversified shape generation · Diversified shape generation | R-Precision: 27.60 $\pm$ 0.39 -> 34.37 $\pm$ 0.09 | Improved |
| Add cyclic loss (ablation) | Diversified shape generation · Diversified shape generation | R-Precision: 34.37 $\pm$ 0.09 -> 40.71 $\pm$ 0.10 | Improved |
| Add two-way cyclic loss to the mixed-feature manipulation method (ablation) | Text-guided shape manipulation · Shape and color manipulation | R-Precision₁: 42.66 $\pm$ 3.77 -> 58.00 $\pm$ 2.82 | Improved |
| Add two-way cyclic loss to the mixed-feature manipulation method (ablation) | Text-guided shape manipulation · Shape and color manipulation | FPD: 35.74 -> 32.03 | Improved |
Closest papers by meaning
- ShapeScaffolder: Structure-Aware 3D Shape Generation from Text (ICCV 2023 · similarity 0.93)
- ShapeWords: Guiding Text-to-Image Synthesis with 3D Shape-Aware Prompts (CVPR 2025 · similarity 0.88)
- PI3D: Efficient Text-to-3D Generation with Pseudo-Image Diffusion (CVPR 2024 · similarity 0.88)
- HyperSDFusion: Bridging Hierarchical Structures in Language and Geometry for Enhanced 3D Text2Shape Generation (CVPR 2024 · similarity 0.88)
- TextField3D: Towards Enhancing Open-Vocabulary 3D Generation with Noisy Text Fields (ICLR 2024 · similarity 0.86)
- ShapeWalk: Compositional Shape Editing Through Language-Guided Chains (CVPR 2024 · 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.
