News: Demo： 简创实验室 ( for Chinese users ) . Although the interface is not currently open, it will be online as soon as possible
News: For a full introduction, please visit 发布 | ControlSpace空间多视角一致性图像生成框架，刷新AIGC设计应用SOTA.
Note: Please don't forget to give us a star if you like this project. Thanks! 😜
ControlSpace: Multi-view consistent image generation framework, updated the AIGC design application SOTA
Institute for AID Lab
The framework consists of two main components: Geometric Consistency and Style Consistency.
Geometric Consistency
The geometric consistency part involves techniques such as iterative alignment strategies and alignment-aware training strategies to establish geometric priors on Nerf/Mip-Nerf models (AGP).
Style Consistency
The style consistency component employs cross-frame non-local attention modules to address frame alignment issues.
The Control Space framework demonstrates a significant advantage in terms of consistency when compared to generating multiple images through ControlNet individually.
Our future work aims to enhance the controllability of 3D scenes based on the Control Space framework.
Email: business@aidlab.tech
