Structural Energy-Guided Sampling for View-Consistent Text-to-3D
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866916913593974784 |
|---|---|
| author | Zhang, Qing Tong, Jinguang Hong, Jie Zhang, Jing Li, Xuesong |
| author_facet | Zhang, Qing Tong, Jinguang Hong, Jie Zhang, Jing Li, Xuesong |
| contents | Text-to-3D generation often suffers from the Janus problem, where objects look correct from the front but collapse into duplicated or distorted geometry from other angles. We attribute this failure to viewpoint bias in 2D diffusion priors, which propagates into 3D optimization. To address this, we propose Structural Energy-Guided Sampling (SEGS), a training-free, plug-and-play framework that enforces multi-view consistency entirely at sampling time. SEGS defines a structural energy in a PCA subspace of intermediate U-Net features and injects its gradients into the denoising trajectory, steering geometry toward the intended viewpoint while preserving appearance fidelity. Integrated seamlessly into SDS/VSD pipelines, SEGS significantly reduces Janus artifacts, achieving improved geometric alignment and viewpoint consistency without retraining or weight modification. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_16917 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Structural Energy-Guided Sampling for View-Consistent Text-to-3D Zhang, Qing Tong, Jinguang Hong, Jie Zhang, Jing Li, Xuesong Computer Vision and Pattern Recognition Text-to-3D generation often suffers from the Janus problem, where objects look correct from the front but collapse into duplicated or distorted geometry from other angles. We attribute this failure to viewpoint bias in 2D diffusion priors, which propagates into 3D optimization. To address this, we propose Structural Energy-Guided Sampling (SEGS), a training-free, plug-and-play framework that enforces multi-view consistency entirely at sampling time. SEGS defines a structural energy in a PCA subspace of intermediate U-Net features and injects its gradients into the denoising trajectory, steering geometry toward the intended viewpoint while preserving appearance fidelity. Integrated seamlessly into SDS/VSD pipelines, SEGS significantly reduces Janus artifacts, achieving improved geometric alignment and viewpoint consistency without retraining or weight modification. |
| title | Structural Energy-Guided Sampling for View-Consistent Text-to-3D |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2508.16917 |