Research map / 3D vision / 3D generation
Debiasing Scores and Prompts of 2D Diffusion for View-consistent Text-to-3D Generation
Susung Hong, Donghoon Ahn, Seungryong 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 score and prompt debiasing to SJC | SJC · 70 user prompts | A-LPIPS (VGG): 0.2054 -> 0.1940 | Improved |
| Add score and prompt debiasing to SJC | SJC · 70 user prompts | A-LPIPS (Alex): 0.1526 -> 0.1445 | Improved |
| Remove contradicting words from user prompts (vs preserving prompts) (ablation) | Debiased SJC · 70 user prompts | A-LPIPS (VGG): 0.1963 -> 0.1940 | Improved |
| Remove contradicting words from user prompts (vs preserving prompts) (ablation) | Debiased SJC · 70 user prompts | A-LPIPS (Alex): 0.1450 -> 0.1445 | Improved |
| Preserve user prompts but apply debiasing (ablation) | SJC · 70 user prompts | A-LPIPS (VGG): 0.2054 -> 0.1963 | Improved |
| Debiasing methods vs baseline in user study | SJC · User study | view consistency: 9.58% -> 90.42% | Improved |
| Debiasing methods vs baseline in user study | SJC · User study | faithfulness: 16.07% -> 83.93% | Improved |
| Debiasing methods vs baseline in user study | SJC · User study | overall quality: 9.67% -> 90.33% | Improved |
| Debiasing methods vs baseline | SJC · 41 prompts with countable faces | success rate: 29.3% -> 68.3% | Improved |
| Prompt debiasing alone, without score debiasing (ablation) | SJC · a smiling cat / panda (qualitative) | extra-ear artifacts | No clear effect |
| Add score debiasing on top of prompt debiasing (ablation) | SJC + prompt debiasing · a smiling cat / panda (qualitative) | extra-ear artifacts | Improved |
| Dynamic clipping of 2D-to-3D scores instead of static clipping (ablation) | Score debiasing · a monkey eating ramen (qualitative) | 3D consistency vs 2D faithfulness | Improved |
Closest papers by meaning
- Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation (ICLR 2024 · similarity 0.92)
- Retrieval-Augmented Score Distillation for Text-to-3D Generation (ICML 2024 · similarity 0.91)
- MVDream: Multi-view Diffusion for 3D Generation (ICLR 2024 · similarity 0.91)
- VP3D: Unleashing 2D Visual Prompt for Text-to-3D Generation (CVPR 2024 · similarity 0.90)
- EfficientDreamer: High-Fidelity and Robust 3D Creation via Orthogonal-view Diffusion Priors (CVPR 2024 · similarity 0.90)
- Taming Mode Collapse in Score Distillation for Text-to-3D Generation (CVPR 2024 · similarity 0.90)
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.
