Research map / 3D vision / Face and shape
ISS: Image as Stepping Stone for Text-Guided 3D Shape Generation
Zhengzhe Liu, Peng Dai, Ruihui Li, XIAOJUAN QI, Chi-Wing Fu
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 stage-1 CLIP image-to-shape alignment (ablation) | ISS · ShapeNet text set | FID: 222.96 -> 124.42 ± 5.11 | Improved |
| Add stage-1 CLIP image-to-shape alignment (ablation) | ISS · ShapeNet text set | Consistency Score: 1.92 ± 2.22 -> 60.0 ± 10.94 | Improved |
| Add test-time text-to-shape alignment (ablation) | ISS · ShapeNet text set | Consistency Score: 29.52 ± 14.86 -> 60.0 ± 10.94 | Improved |
| Add test-time text-to-shape alignment (ablation) | ISS · ShapeNet text set | FID: 202.33 -> 124.42 ± 5.11 | Improved |
| Add the background loss in stage 1 (ablation) | ISS · ShapeNet text set | Consistency Score: 29.45 ± 14.67 -> 60.0 ± 10.94 | Improved |
| Add the background loss in stage 2 (ablation) | ISS · ShapeNet text set | Consistency Score: 31.55 ± 8.87 -> 60.0 ± 10.94 | Improved |
| Add background loss in both stages (ablation) | ISS · ShapeNet text set | Consistency Score: 30.96 ± 15.49 -> 60.0 ± 10.94 | Improved |
| Use the two-stage image-to-shape and text-to-shape alignment instead of the direct CLIP image encoder (ablation) | ISS · ShapeNet text set | FID: 181.88 -> 124.42 ± 5.11 | Improved |
| Use the two-stage image-to-shape and text-to-shape alignment instead of the direct CLIP image encoder (ablation) | ISS · ShapeNet text set | Consistency Score: 20.97 ± 13.59 -> 60.0 ± 10.94 | Improved |
| Generate a shape directly from text with ISS instead of generating an image with LAFITE and reconstructing it with DVR | ISS · ShapeNet text set | FID: 135.01 -> 124.42 ± 5.11 | Improved |
| Generate a shape directly from text with ISS instead of generating an image with LAFITE and reconstructing it with DVR | ISS · ShapeNet text set | Consistency Score: 52.12 ± 11.05 -> 60.0 ± 10.94 | Improved |
| Generate a shape directly from text with ISS instead of using CLIP-Forge | ISS · ShapeNet text set | FID: 162.87 -> 124.42 ± 5.11 | Improved |
Closest papers by meaning
- ShapeWords: Guiding Text-to-Image Synthesis with 3D Shape-Aware Prompts (CVPR 2025 · similarity 0.89)
- ShapeWalk: Compositional Shape Editing Through Language-Guided Chains (CVPR 2024 · similarity 0.86)
- PI3D: Efficient Text-to-3D Generation with Pseudo-Image Diffusion (CVPR 2024 · similarity 0.86)
- HyperSDFusion: Bridging Hierarchical Structures in Language and Geometry for Enhanced 3D Text2Shape Generation (CVPR 2024 · similarity 0.86)
- ViewDiff: 3D-Consistent Image Generation with Text-to-Image Models (CVPR 2024 · similarity 0.86)
- Apply Hierarchical-Chain-of-Generation to Complex Attributes Text-to-3D Generation (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.
