Structural Energy Guidance for View-Consistent Text-to-3D Generation
Researchers have developed a new framework called Structural Energy-Guided Sampling (SEGS) to address the Janus problem in text-to-3D generation. This issue causes inconsistent geometry across different viewpoints. SEGS works by creating a structural energy within the U-Net features and using its gradient during the denoising process, without requiring retraining. Experiments indicate that SEGS can reduce viewpoint inconsistencies by approximately 10% and enhance view consistency scores across various existing text-to-3D models. AI
IMPACT Improves multi-view consistency in 3D generation, potentially enhancing realism and usability of AI-generated 3D assets.