Toward Scene Graph and Layout Guided Complex 3D Scene Generation
Recent advancements in object-centric text-to-3D generation have shown impressive results. However, generating complex 3D scenes remains an open challenge due to the intricate relations between objects. Moreover, existing methods are largely based on score distillation sampling (SDS), which constrains the ability to manipulate multiobjects with specific interactions. Addressing these critical yet underexplored issues, we present a novel framework of Scene Graph and Layout Guided 3D Scene Generation (GraLa3D). Given a text prompt describing a complex 3D scene, GraLa3D utilizes LLM to model the scene using a scene graph representation with layout bounding box information. GraLa3D uniquely constructs the scene graph with single-object nodes and composite super-nodes. In addition to constraining 3D generation within the desirable layout, a major contribution lies in the modeling of interactions between objects in a super-node, while alleviating appearance leakage across objects within such nodes. Our experiments confirm that GraLa3D overcomes the above limitations and generates complex 3D scenes closely aligned with text prompts.
Code (0)
등록된 구현이 없습니다.
Tasks
3D GenerationScene GenerationText to 3DSimilar Papers 제목 키워드 기반
LAYOUTDREAMER: Physics-guided Layout for Text-to-3D Compositional Scene Generation
Recently, the field of text-guided 3D scene generation has garnered significant attention. High-quality generation that aligns with physical realism and high controllability is crucial for practical 3D scene applications…
3DGSScene GenerationText to 3DImaginarium: Vision-guided High-Quality 3D Scene Layout Generation
Generating artistic and coherent 3D scene layouts is crucial in digital content creation. Traditional optimization-based methods are often constrained by cumbersome manual rules, while deep generative models face challen…
Image GenerationImage Synthesis with Graph Conditioning: CLIP-Guided Diffusion Models for Scene Graphs
Advancements in generative models have sparked significant interest in generating images while adhering to specific structural guidelines. Scene graph to image generation is one such task of generating images which are c…
Image GenerationExploiting Relationship for Complex-scene Image Generation
The significant progress on Generative Adversarial Networks (GANs) has facilitated realistic single-object image generation based on language input. However, complex-scene generation (with various interactions among mult…
Image GenerationScene GenerationGALA3D: Towards Text-to-3D Complex Scene Generation via Layout-guided Generative Gaussian Splatting
We present GALA3D, generative 3D GAussians with LAyout-guided control, for effective compositional text-to-3D generation. We first utilize large language models (LLMs) to generate the initial layout and introduce a layou…
3D GenerationScene GenerationText to 3D