The generation of realistic textures in 3D modeling has always been a research hotspot. Recently, the University of Alberta, the University of Toronto, and Huawei's Noah's Ark Laboratory collaborated to develop a new method called TexGen. This method can generate corresponding 3D textures based on text descriptions provided by users, and its effect is significantly better than existing ones. There is technology. The core of TexGen lies in the attention-guided multi-view sampling strategy and noise resampling technology. The former ensures the consistency of textures under different viewing angles, while the latter protects texture details and improves texture editing capabilities. The research results have made significant progress in terms of texture quality, viewing angle consistency and detail richness.
In the field of 3D modeling, how to generate realistic textures has always been a difficult problem. Recently, researchers from the University of Alberta, the University of Toronto, and Huawei's Noah's Ark Laboratory jointly launched a new method-TexGen. This technology can generate 3D textures corresponding to the user's text description, and its effect is significantly better than existing similar technologies.
TexGen also introduces an attention-guided multi-view sampling strategy, through which the appearance information of the texture can be broadcast between different views, thereby ensuring that the generated texture is consistent at each view.
In addition, to preserve texture details, the research team also developed a noise resampling technique, which can help estimate noise and generate suitable input for subsequent denoising steps. In this way, TexGen not only achieves a breakthrough in texture generation, but also excels in texture editing, able to adjust while retaining the original identity.
After extensive qualitative and quantitative evaluations, the research team found that TexGen performed extremely well in generating texture quality, perspective consistency, and rich appearance details of diverse 3D objects, surpassing the current state-of-the-art methods.
Below is a comparison of TexGen with TEXTure, Text2Tex, Fantasia3D and ProlificDreamer
Here are some effects to show:
Highlight:
TexGen can generate high-quality 3D textures from text descriptions, solving the seams and over-smoothing problems in traditional methods.
? Adopt a multi-view sampling and resampling framework to ensure texture consistency between each view.
?️ The new noise resampling technology makes TexGen also perform well in texture editing and can retain the original style.
The emergence of TexGen has brought new possibilities to the field of 3D modeling. Its efficient text-driven texture generation capabilities and precise control of details will greatly improve modeling efficiency and texture quality. In the future, we look forward to TexGen’s performance in more application scenarios.