Running summarizing tools on a new article
This is the first time this article is requested and our AI summarizing tools have never been run on it. We can run our tools now if you click on the button "Run" donw the page but first make sure that it is the right article.
DreamPolisher: Towards High-Quality Text-to-3D Generation via Geometric Diffusion
Yuanze Lin, Ronald Clark, Philip Torr
Abstract: We present DreamPolisher, a novel Gaussian Splatting based method with geometric guidance, tailored to learn cross-view consistency and intricate detail from textual descriptions. While recent progress on text-to-3D generation methods have been promising, prevailing methods often fail to ensure view-consistency and textural richness. This problem becomes particularly noticeable for methods that work with text input alone. To address this, we propose a two-stage Gaussian Splatting based approach that enforces geometric consistency among views. Initially, a coarse 3D generation undergoes refinement via geometric optimization. Subsequently, we use a ControlNet driven refiner coupled with the geometric consistency term to improve both texture fidelity and overall consistency of the generated 3D asset. Empirical evaluations across diverse textual prompts spanning various object categories demonstrate the efficacy of DreamPolisher in generating consistent and realistic 3D objects, aligning closely with the semantics of the textual instructions.
⚠The license of this specific paper does not allow us to build upon its content and the summarizing tools will be run using the paper metadata rather than the full article. However, it still does a good job, and you can also try our tools on papers with more open licenses.