MimiCAT 3D Pose Transfer Model

MimiCAT is a cutting-edge artificial intelligence framework designed for category-free 3D pose transfer. It was developed through a collaborative effort involving researchers from leading institutions in Singapore and Hong Kong. The project represents a significant advancement in computer vision, specifically addressing the challenge of transferring skeletal animations between characters of vastly different morphologies, such as from a humanoid to a quadruped.

The development and technical origin of MimiCAT are rooted in the academic research sectors of the National University of Singapore and the Chinese University of Hong Kong. As a research project published for CVPR 2026, its 'manufacturing' or development environment is distributed across these academic laboratories. The tool utilizes a cascade-transformer architecture to estimate soft keypoint correspondences, allowing for highly flexible 3D modeling and animation.

Ownership of the project remains with the academic institutions and the individual researchers who authored the framework. It is typically released under open-source licenses for the global research community to use and iterate upon. The project is supported by prominent research bodies in Asia, including Singapore's Agency for Science, Technology and Research (A*STAR).