3D Generation
A technology that uses AI models to generate three-dimensional objects or scenes, which can be realized through web code (Three.js/WebGL) or traditional 3D asset files.
Timeline (last 90 days)
SUPRA-TEX is a framework for generating material-bearing, relightable 3D assets from a single image; by decoupling visible material estimation, surface alignment, and observation-guided 3D completion, it reaches SOTA level when trained only on public datasets.
SUPRA-TEX validates a design principle: decouple the generation task from the perception task, and use mature capabilities from the 2D domain to feed back into 3D generation.
In a manually annotated pool of 16 candidates, using oracle selection can raise the full-fact success rate from 26.7% to 90.0%, an improvement of more than 3 times.
In the context of 3D generation, Factual Correctness refers to whether the generated content satisfies the explicit or implicit factual constraints in the task specification, including object category, attributes, quantity, spatial relationships, size proportions, physical plausibility, etc., and cannot be judged by rendering quality alone.
Evaluation standards in the 3D generation field have long favored visual quality, but visual fidelity and factual correctness are two different things; a 3D scene may look impeccable yet have hard flaws in object attributes, spatial relationships, and numerical specifications.
SVG and interactive 3D content generation tasks require models to simultaneously possess three capabilities: spatial geometry reasoning, object structure understanding, and precise code generation.
All Facts (7)
3D生成面临的挑战远比2D复杂，生成结果需要满足拓扑合理性、物理正确性和渲染效率等多重约束
90%UnverifiedSUPRA-TEX validates a design principle: decouple the generation task from the perception task, and use mature capabilities from the 2D domain to feed back into 3D generation.
50%UnverifiedSUPRA-TEX is a framework for generating material-bearing, relightable 3D assets from a single image; by decoupling visible material estimation, surface alignment, and observation-guided 3D completion, it reaches SOTA level when trained only on public datasets.
50%UnverifiedIn the context of 3D generation, Factual Correctness refers to whether the generated content satisfies the explicit or implicit factual constraints in the task specification, including object category, attributes, quantity, spatial relationships, size proportions, physical plausibility, etc., and cannot be judged by rendering quality alone.
50%UnverifiedIn a manually annotated pool of 16 candidates, using oracle selection can raise the full-fact success rate from 26.7% to 90.0%, an improvement of more than 3 times.
50%UnverifiedEvaluation standards in the 3D generation field have long favored visual quality, but visual fidelity and factual correctness are two different things; a 3D scene may look impeccable yet have hard flaws in object attributes, spatial relationships, and numerical specifications.
50%UnverifiedSVG and interactive 3D content generation tasks require models to simultaneously possess three capabilities: spatial geometry reasoning, object structure understanding, and precise code generation.
50%