aigc3d
Target pose and multi-view rendering
Very excellent work! I'm wondering if there's an update to this part of the work for getting the T pose and then transforming to the target pose and rendering to get the multi-view result？Thank you!
While we currently have no plans to implement this specific functionality, the approach is relatively straightforward to realize. The implementation would involve three key steps: First, modifying the existing code to generate target Gaussian splats (GS) for the desired poses. Second, transforming these GS representations back to their original coordinate space after saving them as Gaussian PLY files. Finally, one could render multi-view images using a virtual camera with self-rotation capability.
